Chimera Software User GuideTutorials & Model DemosModel DemosModel Demo: YOLOv8 Pose Estimation

Model Demo: YOLOv8 Pose Estimation

NOTE: The Jupyter Notebook below is included in the Chimera SDK and can be run interactively by running the following CLI command:

$ quadric sdk notebook

From the Jupyter Notebook window in your browser, select the notebook named /quadric/sdk-cli/examples/models/yolo/yolov8_pose/yolov8_pose.ipynb.

YOLOv8 Pose Estimation

YOLOv8 Pose Estimation has been released from Ultralytics. Details can be found here.

Environment Setup

!pip3 install -r ../../../requirements.txt -q

[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv[0m[33m
[0m

import gc
import onnx
import onnxsim
from pathlib import Path

from examples.models.zoo.zoo_utils import onnx_check_and_simplify

Model Selection

YOLOv8 supports pretrained models with various sizes. The following table is excerpted from YOLOv8 page

Model	size (pixels)	mAPpose 50-95	mAPpose 50	Speed CPU ONNX (ms)	params (M)	FLOPs (B)
YOLOv8n-pose	640	50.4	80.1	131.8	3.3	9.2
YOLOv8s-pose	640	60.0	86.2	233.2	11.6	30.2
YOLOv8m-pose	640	65.0	88.8	456.3	26.4	81.0
YOLOv8l-pose	640	67.6	90.0	784.5	44.4	168.6
YOLOv8x-pose	640	69.2	90.2	1607.1	69.4	263.2
YOLOv8x-pose-p6	1280	71.6	91.2	4088.7	99.1	1066.4

Here, we are going to use yolov8n-pose to demonstrate the Chimera capability on YOLOv8 Pose Estimation. In this notebook, yolov8s-pose can be experimented. Please contact the Quadric sales team for larger models.

MODEL_NAME = "yolov8n-pose"

Model Generation and ONNX Export

from ultralytics import YOLO
from tvm.contrib.epu.chimera_job.constants import DEFAULT_ONNX_OPSET

## Load a model
model = YOLO(f"{MODEL_NAME}.pt")  # load an official model

## Export the model
model.export(format="onnx", opset=DEFAULT_ONNX_OPSET)

Downloading https://github.com/ultralytics/assets/releases/download/v8.3.0/yolov8n-pose.pt to 'yolov8n-pose.pt'...


100%|█████████████████████████████████████| 6.52M/6.52M [00:00<00:00, 65.3MB/s]

Ultralytics 8.3.166 🚀 Python-3.10.12 torch-2.6.0+cpu CPU (AMD Ryzen 9 9950X 16-Core Processor)





YOLOv8n-pose summary (fused): 81 layers, 3,289,964 parameters, 0 gradients, 9.2 GFLOPs

[34m[1mPyTorch:[0m starting from 'yolov8n-pose.pt' with input shape (1, 3, 640, 640) BCHW and output shape(s) (1, 56, 8400) (6.5 MB)
[31m[1mrequirements:[0m Ultralytics requirement ['onnxslim>=0.1.59'] not found, attempting AutoUpdate...
WARNING ⚠️ Retry 1/2 failed: Command 'pip install --no-cache-dir "onnxslim>=0.1.59" ' returned non-zero exit status 127.
WARNING ⚠️ Retry 2/2 failed: Command 'pip install --no-cache-dir "onnxslim>=0.1.59" ' returned non-zero exit status 127.
WARNING ⚠️ [31m[1mrequirements:[0m ❌ Command 'pip install --no-cache-dir "onnxslim>=0.1.59" ' returned non-zero exit status 127.

[34m[1mONNX:[0m starting export with onnx 1.16.2 opset 16...
WARNING ⚠️ [34m[1mONNX:[0m simplifier failure: No module named 'onnxslim'
[34m[1mONNX:[0m export success ✅ 2.3s, saved as 'yolov8n-pose.onnx' (12.9 MB)

Export complete (2.4s)
Results saved to [1m/quadric/sdk-cli/examples/models/yolo/yolov8_pose[0m
Predict:         yolo predict task=pose model=yolov8n-pose.onnx imgsz=640  
Validate:        yolo val task=pose model=yolov8n-pose.onnx imgsz=640 data=/usr/src/app/ultralytics/datasets/coco-pose.yaml  
Visualize:       https://netron.app





'yolov8n-pose.onnx'

Split ONNX

When the above cell is successfully executed, yolov8n-pose.onnx is generated in the current directory.

This command cell uses this onnx file to split to a backbone and a head.

from tvm.contrib.epu.onnx_util import cut_onnx

onnx_file = f"{MODEL_NAME}.onnx"

input_edges = ["images"]
output_edges = ["output0"]
split_edges = [
    "/model.22/cv2.0/cv2.0.1/act/Mul_output_0",
    "/model.22/cv3.0/cv3.0.1/act/Mul_output_0",
    "/model.22/cv2.1/cv2.1.1/act/Mul_output_0",
    "/model.22/cv3.1/cv3.1.1/act/Mul_output_0",
    "/model.22/cv2.2/cv2.2.1/act/Mul_output_0",
    "/model.22/cv3.2/cv3.2.1/act/Mul_output_0",
    "/model.22/cv4.0/cv4.0.1/act/Mul_output_0",
    "/model.22/cv4.1/cv4.1.1/act/Mul_output_0",
    "/model.22/cv4.2/cv4.2.1/act/Mul_output_0",
]

model = onnx.load(onnx_file)
backbone_onnx_file = f"{MODEL_NAME}-backbone.onnx"
head_onnx_file = f"{MODEL_NAME}-head.onnx"

## Extract the backbone.
sub_graph = cut_onnx(model, input_edges, split_edges)
sub_graph = onnx_check_and_simplify(sub_graph)
onnx.save(sub_graph, backbone_onnx_file)
print(f"Backbone onnx is saved as {backbone_onnx_file}")

## Extract the head.
sub_graph = cut_onnx(model, split_edges, output_edges)
sub_graph = onnx_check_and_simplify(sub_graph)
onnx.save(sub_graph, head_onnx_file)
print(f"Head onnx is saved as {head_onnx_file}")

Backbone onnx is saved as yolov8n-pose-backbone.onnx
Head onnx is saved as yolov8n-pose-head.onnx

Quantization

from torch.utils.data import Subset
from torchvision.transforms import Compose, Normalize, ToTensor
from sdk_cli.utils.transforms import ResizePad
from sdk_cli.lib.quantize import (
    QuantizedONNXModel,
    quantize_onnx_model,
)
from sdk_cli.utils.datasets import QuadricCalibration

dataset_input_size = (640, 640)

dataset_mean, dataset_std = (0, 0, 0), (1, 1, 1)

transforms = Compose(
    [
        ResizePad(dataset_input_size),
        ToTensor(),
        Normalize(dataset_mean, dataset_std),
    ]
)

## Path to directory containing data to use for for numerical range calibration during quantization
## Data is used also used to compare accuracy of fp32 and int8 models
dataset = QuadricCalibration.Dataset(transform=transforms)

## NOTE: `coco-like` is the 0th index target for `QuadricCalibration.Dataset`
coco_like_data_indices = [index for index, target in enumerate(dataset.targets) if target == 0]
coco_like_subset_of_dataset = Subset(dataset, coco_like_data_indices)

quantized_onnx_model: QuantizedONNXModel = quantize_onnx_model(
    backbone_onnx_file,
    dataset,
    asymmetric_activation=True,
)
print(
    f"quantized onnx: {quantized_onnx_model.model_path}, tranges file: {quantized_onnx_model.tensor_ranges_path}"
)
gc.collect();

2026-07-18 12:09 - INFO - sdk - quantize - ONNX model shapes inferred.
2026-07-18 12:09 - DEBUG - sdk - quantize - Forcing node types: []
2026-07-18 12:09 - DEBUG - sdk - quantize - ONNX Node types excluded from quantization: ['Softmax', 'Sigmoid', 'QuadricCustomOp']
2026-07-18 12:09 - DEBUG - sdk - quantize - ONNX Node names excluded from quantization: ['/model.22/cv4.1/cv4.1.0/act/Mul', '/model.4/m.0/cv2/act/Mul', '/model.22/cv2.2/cv2.2.0/act/Sigmoid', '/model.4/cv2/act/Mul', '/model.19/act/Sigmoid', '/model.9/cv1/act/Sigmoid', '/model.12/m.0/cv1/act/Sigmoid', '/model.8/cv1/act/Sigmoid', '/model.7/act/Sigmoid', '/model.18/m.0/cv2/act/Mul', '/model.21/cv2/act/Sigmoid', '/model.12/m.0/cv2/act/Sigmoid', '/model.15/cv1/act/Mul', '/model.21/cv2/act/Mul', '/model.5/act/Sigmoid', '/model.18/cv2/act/Mul', '/model.6/m.0/cv2/act/Sigmoid', '/model.12/cv1/act/Mul', '/model.8/cv2/act/Sigmoid', '/model.6/m.1/cv2/act/Sigmoid', '/model.22/cv3.1/cv3.1.0/act/Sigmoid', '/model.4/m.1/cv2/act/Mul', '/model.8/m.0/cv2/act/Mul', '/model.15/m.0/cv1/act/Mul', '/model.15/m.0/cv2/act/Sigmoid', '/model.6/cv2/act/Mul', '/model.1/act/Sigmoid', '/model.19/act/Mul', '/model.22/cv4.1/cv4.1.1/act/Mul', '/model.8/m.0/cv2/act/Sigmoid', '/model.7/act/Mul', '/model.6/m.1/cv2/act/Mul', '/model.4/m.1/cv1/act/Mul', '/model.15/m.0/cv1/act/Sigmoid', '/model.8/cv1/act/Mul', '/model.22/cv3.0/cv3.0.0/act/Sigmoid', '/model.4/m.0/cv1/act/Sigmoid', '/model.15/m.0/cv2/act/Mul', '/model.6/m.0/cv1/act/Sigmoid', '/model.22/cv2.2/cv2.2.0/act/Mul', '/model.3/act/Sigmoid', '/model.8/cv2/act/Mul', '/model.21/cv1/act/Sigmoid', '/model.12/m.0/cv2/act/Mul', '/model.1/act/Mul', '/model.9/cv2/act/Mul', '/model.4/cv1/act/Mul', '/model.12/cv2/act/Mul', '/model.12/m.0/cv1/act/Mul', '/model.4/m.0/cv1/act/Mul', '/model.4/cv2/act/Sigmoid', '/model.4/cv1/act/Sigmoid', '/model.22/cv4.0/cv4.0.0/act/Sigmoid', '/model.18/cv2/act/Sigmoid', '/model.22/cv2.1/cv2.1.1/act/Mul', '/model.0/act/Mul', '/model.21/m.0/cv1/act/Mul', '/model.22/cv4.1/cv4.1.0/act/Sigmoid', '/model.22/cv4.2/cv4.2.1/act/Sigmoid', '/model.22/cv2.0/cv2.0.1/act/Sigmoid', '/model.4/m.1/cv1/act/Sigmoid', '/model.22/cv2.1/cv2.1.0/act/Sigmoid', '/model.22/cv4.2/cv4.2.0/act/Mul', '/model.22/cv4.0/cv4.0.1/act/Sigmoid', '/model.16/act/Sigmoid', '/model.22/cv2.0/cv2.0.1/act/Mul', '/model.15/cv2/act/Mul', '/model.22/cv3.2/cv3.2.0/act/Mul', '/model.0/act/Sigmoid', '/model.22/cv4.1/cv4.1.1/act/Sigmoid', '/model.2/m.0/cv2/act/Mul', '/model.9/cv1/act/Mul', '/model.8/m.0/cv1/act/Sigmoid', '/model.15/cv2/act/Sigmoid', '/model.22/cv3.2/cv3.2.0/act/Sigmoid', '/model.6/m.1/cv1/act/Mul', '/model.8/m.0/cv1/act/Mul', '/model.6/cv1/act/Sigmoid', '/model.2/cv1/act/Sigmoid', '/model.2/m.0/cv1/act/Sigmoid', '/model.16/act/Mul', '/model.18/cv1/act/Mul', '/model.22/cv4.0/cv4.0.0/act/Mul', '/model.22/cv3.1/cv3.1.1/act/Mul', '/model.12/cv1/act/Sigmoid', '/model.3/act/Mul', '/model.2/m.0/cv1/act/Mul', '/model.21/m.0/cv1/act/Sigmoid', '/model.22/cv4.2/cv4.2.0/act/Sigmoid', '/model.22/cv2.2/cv2.2.1/act/Mul', '/model.2/cv2/act/Mul', '/model.21/cv1/act/Mul', '/model.22/cv4.0/cv4.0.1/act/Mul', '/model.22/cv4.2/cv4.2.1/act/Mul', '/model.5/act/Mul', '/model.18/cv1/act/Sigmoid', '/model.2/cv2/act/Sigmoid', '/model.21/m.0/cv2/act/Sigmoid', '/model.22/cv2.0/cv2.0.0/act/Mul', '/model.15/cv1/act/Sigmoid', '/model.22/cv2.1/cv2.1.1/act/Sigmoid', '/model.6/m.0/cv2/act/Mul', '/model.6/m.1/cv1/act/Sigmoid', '/model.22/cv3.0/cv3.0.1/act/Mul', '/model.22/cv3.0/cv3.0.0/act/Mul', '/model.18/m.0/cv1/act/Sigmoid', '/model.22/cv3.1/cv3.1.0/act/Mul', '/model.9/cv2/act/Sigmoid', '/model.6/cv2/act/Sigmoid', '/model.18/m.0/cv1/act/Mul', '/model.22/cv2.0/cv2.0.0/act/Sigmoid', '/model.4/m.0/cv2/act/Sigmoid', '/model.22/cv3.2/cv3.2.1/act/Mul', '/model.12/cv2/act/Sigmoid', '/model.18/m.0/cv2/act/Sigmoid', '/model.2/cv1/act/Mul', '/model.22/cv3.2/cv3.2.1/act/Sigmoid', '/model.6/cv1/act/Mul', '/model.22/cv2.2/cv2.2.1/act/Sigmoid', '/model.22/cv3.0/cv3.0.1/act/Sigmoid', '/model.6/m.0/cv1/act/Mul', '/model.21/m.0/cv2/act/Mul', '/model.2/m.0/cv2/act/Sigmoid', '/model.22/cv3.1/cv3.1.1/act/Sigmoid', '/model.4/m.1/cv2/act/Sigmoid', '/model.22/cv2.1/cv2.1.0/act/Mul']
2026-07-18 12:09 - INFO - sdk - quantize - Starting quantization...
WARNING:root:Please use QuantFormat.QDQ for activation type QInt8 and weight type QInt8. Or it will lead to bad performance on x64.
2026-07-18 12:09 - INFO - sdk - quantize - Quantization completed! Quantized model saved to /quadric/sdk-cli/examples/models/yolo/yolov8_pose/yolov8n-pose-backbone_OpSet16_optimized_asym_int8_q.onnx
2026-07-18 12:09 - INFO - sdk - quantize - ONNX full precision model size: 12.53MB
2026-07-18 12:09 - INFO - sdk - quantize - ONNX quantized model size: 3.26MB
2026-07-18 12:09 - INFO - sdk - quantize - ONNX model shapes inferred.
2026-07-18 12:09 - INFO - sdk - quantize - ONNX Model with well-defined shapes has been saved at `/quadric/sdk-cli/examples/models/yolo/yolov8_pose/yolov8n-pose-backbone_OpSet16_optimized_asym_int8_q_shaped.onnx`.
2026-07-18 12:09 - DEBUG - sdk - quantize - Checking for FLOAT/FLOAT16 types...
2026-07-18 12:09 - INFO - sdk - quantize - Checking for remaining FLOAT/FLOAT16 types.
2026-07-18 12:09 - INFO - sdk - quantize - Model still has FLOAT/FLOAT16 types after quantization. Creating ranges for floating point tensors using calibration data...
2026-07-18 12:10 - INFO - sdk - quantize - Saved computed tensor ranges to /quadric/sdk-cli/examples/models/yolo/yolov8_pose/yolov8n-pose-backbone_OpSet16_optimized_asym_int8_q_shaped.tranges.
2026-07-18 12:10 - INFO - sdk - quantize - 
╒═══════════════════════════════════════════════════════════════════════════════════════════════════════════════════╤══════════════════════════════════════════════════════════════════════════════════════════════════════════════════════╕
│ Quantized ONNX Model                                                                                              │ Tensor Ranges File                                                                                                   │
╞═══════════════════════════════════════════════════════════════════════════════════════════════════════════════════╪══════════════════════════════════════════════════════════════════════════════════════════════════════════════════════╡
│ /quadric/sdk-cli/examples/models/yolo/yolov8_pose/yolov8n-pose-backbone_OpSet16_optimized_asym_int8_q_shaped.onnx │ /quadric/sdk-cli/examples/models/yolo/yolov8_pose/yolov8n-pose-backbone_OpSet16_optimized_asym_int8_q_shaped.tranges │
╘═══════════════════════════════════════════════════════════════════════════════════════════════════════════════════╧══════════════════════════════════════════════════════════════════════════════════════════════════════════════════════╛


quantized onnx: /quadric/sdk-cli/examples/models/yolo/yolov8_pose/yolov8n-pose-backbone_OpSet16_optimized_asym_int8_q_shaped.onnx, tranges file: /quadric/sdk-cli/examples/models/yolo/yolov8_pose/yolov8n-pose-backbone_OpSet16_optimized_asym_int8_q_shaped.tranges

Compilation

from tvm.contrib.epu.chimera_job.chimera_job import ChimeraJob


cgc_job = ChimeraJob(
    model_p=str(quantized_onnx_model.model_path),
)
cgc_job.compile(quiet=True)
print(cgc_job)

╒═════════════════════╤═══════════════════════════════════════════════════════════════════════════════════════════════════════════════════╕
│ Module Name         │ yolov8n_pose_backbone_OpSet16_optimized_asym_int8_q_shaped_QC_U_1d7_16MB_4kB_128GBps_128GBps_16_OFF_x1_x1         │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ ONNX File           │ /quadric/sdk-cli/examples/models/yolo/yolov8_pose/yolov8n-pose-backbone_OpSet16_optimized_asym_int8_q_shaped.onnx │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Product Target      │ QC-U                                                                                                              │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Number of Cores     │ 1                                                                                                                 │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ ISS Clock Frequency │ 1.700                                                                                                             │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ L2M Size            │ 16MB                                                                                                              │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ LRM Size            │ 4kB                                                                                                               │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ External Read BW    │ 128GBps                                                                                                           │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ External Write BW   │ 128GBps                                                                                                           │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ MACS per PE         │ 16                                                                                                                │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Max L2M             │ 5.892MB                                                                                                           │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Max LRM             │ 1.750kB                                                                                                           │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Max Temp Ext Bytes  │ 0.000MB                                                                                                           │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Network GMACs       │ 4.532                                                                                                             │
╘═════════════════════╧═══════════════════════════════════════════════════════════════════════════════════════════════════════════════════╛

╒════╤════════╤══════════════════════════════════════════╤══════════════════╤══════════════════════════╤═══════╕
│    │ Type   │ Name                                     │ shape            │ type                     │ mse   │
╞════╪════════╪══════════════════════════════════════════╪══════════════════╪══════════════════════════╪═══════╡
│  0 │ Input  │ images                                   │ [1, 3, 640, 640] │ tensor[FixedPoint32<27>] │ n/a   │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  1 │ Output │ /model.22/cv2.0/cv2.0.1/act/Mul_output_0 │ [1, 64, 80, 80]  │ tensor[FixedPoint32<25>] │ n/a   │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  2 │ Output │ /model.22/cv3.0/cv3.0.1/act/Mul_output_0 │ [1, 64, 80, 80]  │ tensor[FixedPoint32<25>] │ n/a   │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  3 │ Output │ /model.22/cv2.1/cv2.1.1/act/Mul_output_0 │ [1, 64, 40, 40]  │ tensor[FixedPoint32<25>] │ n/a   │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  4 │ Output │ /model.22/cv3.1/cv3.1.1/act/Mul_output_0 │ [1, 64, 40, 40]  │ tensor[FixedPoint32<25>] │ n/a   │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  5 │ Output │ /model.22/cv2.2/cv2.2.1/act/Mul_output_0 │ [1, 64, 20, 20]  │ tensor[FixedPoint32<25>] │ n/a   │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  6 │ Output │ /model.22/cv3.2/cv3.2.1/act/Mul_output_0 │ [1, 64, 20, 20]  │ tensor[FixedPoint32<25>] │ n/a   │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  7 │ Output │ /model.22/cv4.0/cv4.0.1/act/Mul_output_0 │ [1, 51, 80, 80]  │ tensor[FixedPoint32<26>] │ n/a   │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  8 │ Output │ /model.22/cv4.1/cv4.1.1/act/Mul_output_0 │ [1, 51, 40, 40]  │ tensor[FixedPoint32<27>] │ n/a   │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  9 │ Output │ /model.22/cv4.2/cv4.2.1/act/Mul_output_0 │ [1, 51, 20, 20]  │ tensor[FixedPoint32<27>] │ n/a   │
╘════╧════════╧══════════════════════════════════════════╧══════════════════╧══════════════════════════╧═══════╛

Demo

Inference as Batch

ChimeraJob supports batch execution which invokes specified number of threads to execute inference in parallel, which shortens ISS inference.

Inference

import numpy as np
from onnxruntime import InferenceSession
from PIL import Image
from sdk_cli.lib.inference import InferenceEngine, batch_inference

all_image_paths = [
    "../../../common/calibration/face/daniel_maverick.png",
    "../../../common/calibration/face/sales_squad_jani.jpg",
    "../../../common/calibration/coco-like/33823288584_1d21cf0a26_k.jpeg",
]
NUM_IMAGES = len(all_image_paths)

all_images = []
for image_path in all_image_paths:
    image = Image.open(image_path)
    transformed_image = transforms(image)
    all_images.append(np.expand_dims(transformed_image.numpy(), axis=0).astype(np.float32))

engines = {
    InferenceEngine.CHIMERA_ORT_INT8: cgc_job,
    InferenceEngine.CHIMERA_ISS_INT8: cgc_job,
}

head_onnx_model = onnx.load(head_onnx_file)
head_session = InferenceSession(head_onnx_model.SerializeToString())

outputs_per_inference_engine = {}
THREADS = min(NUM_IMAGES, 6)
for inference_engine, engine in engines.items():
    all_outputs = batch_inference(
        inference_engine,
        engine,
        all_images,
        head_session=head_session,
        threads=THREADS,
    )
    outputs_per_inference_engine[inference_engine] = all_outputs

2026-07-18 12:15 - WARNING - epu - chimera_job - ORT is not threadsafe -- forcing single threaded batch execution
100%|████████████████████████████████████████████| 3/3 [00:00<00:00,  3.91it/s]
Processing: 100%|████████████████████████████████| 3/3 [00:58<00:00, 19.37s/it]

Run Statistics for PoseResnet Backbone

print(cgc_job)
cgc_job.plot_run_statistics()

╒═════════════════════╤═══════════════════════════════════════════════════════════════════════════════════════════════════════════════════╕
│ Module Name         │ yolov8n_pose_backbone_OpSet16_optimized_asym_int8_q_shaped_QC_U_1d7_16MB_4kB_128GBps_128GBps_16_OFF_x1_x1         │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ ONNX File           │ /quadric/sdk-cli/examples/models/yolo/yolov8_pose/yolov8n-pose-backbone_OpSet16_optimized_asym_int8_q_shaped.onnx │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Product Target      │ QC-U                                                                                                              │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Number of Cores     │ 1                                                                                                                 │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ ISS Clock Frequency │ 1.700                                                                                                             │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ L2M Size            │ 16MB                                                                                                              │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ LRM Size            │ 4kB                                                                                                               │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ External Read BW    │ 128GBps                                                                                                           │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ External Write BW   │ 128GBps                                                                                                           │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ MACS per PE         │ 16                                                                                                                │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Max L2M             │ 5.892MB                                                                                                           │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Max LRM             │ 1.750kB                                                                                                           │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Max Temp Ext Bytes  │ 0.000MB                                                                                                           │
├─────────────────────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ Network GMACs       │ 4.532                                                                                                             │
╘═════════════════════╧═══════════════════════════════════════════════════════════════════════════════════════════════════════════════════╛

╒════╤════════╤══════════════════════════════════════════╤══════════════════╤══════════════════════════╤═══════╕
│    │ Type   │ Name                                     │ shape            │ type                     │ mse   │
╞════╪════════╪══════════════════════════════════════════╪══════════════════╪══════════════════════════╪═══════╡
│  0 │ Input  │ images                                   │ [1, 3, 640, 640] │ tensor[FixedPoint32<27>] │ n/a   │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  1 │ Output │ /model.22/cv2.0/cv2.0.1/act/Mul_output_0 │ [1, 64, 80, 80]  │ tensor[FixedPoint32<25>] │ 0.091 │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  2 │ Output │ /model.22/cv3.0/cv3.0.1/act/Mul_output_0 │ [1, 64, 80, 80]  │ tensor[FixedPoint32<25>] │ 0.026 │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  3 │ Output │ /model.22/cv2.1/cv2.1.1/act/Mul_output_0 │ [1, 64, 40, 40]  │ tensor[FixedPoint32<25>] │ 0.075 │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  4 │ Output │ /model.22/cv3.1/cv3.1.1/act/Mul_output_0 │ [1, 64, 40, 40]  │ tensor[FixedPoint32<25>] │ 0.036 │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  5 │ Output │ /model.22/cv2.2/cv2.2.1/act/Mul_output_0 │ [1, 64, 20, 20]  │ tensor[FixedPoint32<25>] │ 0.051 │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  6 │ Output │ /model.22/cv3.2/cv3.2.1/act/Mul_output_0 │ [1, 64, 20, 20]  │ tensor[FixedPoint32<25>] │ 0.040 │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  7 │ Output │ /model.22/cv4.0/cv4.0.1/act/Mul_output_0 │ [1, 51, 80, 80]  │ tensor[FixedPoint32<26>] │ 0.027 │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  8 │ Output │ /model.22/cv4.1/cv4.1.1/act/Mul_output_0 │ [1, 51, 40, 40]  │ tensor[FixedPoint32<27>] │ 0.012 │
├────┼────────┼──────────────────────────────────────────┼──────────────────┼──────────────────────────┼───────┤
│  9 │ Output │ /model.22/cv4.2/cv4.2.1/act/Mul_output_0 │ [1, 51, 20, 20]  │ tensor[FixedPoint32<27>] │ 0.006 │
╘════╧════════╧══════════════════════════════════════════╧══════════════════╧══════════════════════════╧═══════╛

Post-ISS Report 1.7 GHz ***
Fully placed-and-routed gate simulation: 
╒══════════════════════════════════╤═════════╕
│ Latency (ms)                     │ 1.09    │
├──────────────────────────────────┼─────────┤
│ FPS                              │ 914.92  │
├──────────────────────────────────┼─────────┤
│ Average Power @ 3nm SSGNP (mW)   │ 2033.06 │
├──────────────────────────────────┼─────────┤
│ FPS per Watt @ 3nm SSGNP (FPS/W) │ 450.02  │
├──────────────────────────────────┼─────────┤
│ Ext Rd Bytes (MB)                │ 7.89    │
├──────────────────────────────────┼─────────┤
│ Ext Wr Bytes (MB)                │ 5.74    │
├──────────────────────────────────┼─────────┤
│ Avg Ext Rd BW (GBps)             │ 7.05    │
├──────────────────────────────────┼─────────┤
│ Avg Ext Wr BW (GBps)             │ 5.12    │
├──────────────────────────────────┼─────────┤
│ MAC Utilization                  │ 14.89%  │
╘══════════════════════════════════╧═════════╛
*** Data generated using 7nm SSGNP gatesim and scaled to 3nm

[SDK-CLI] : TotalCycles: 1,858,080
[SDK-CLI] : Executions/second: 914.92

compute      : ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 599.352K
data_array   : ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 215.267K
mac          : ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 630.826K
data_external: ▇▇▇▇▇▇▇▇▇▇ 138.431K
data_ocm     : ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 272.166K

for more information check run directory: /quadric/sdk-cli/examples/models/yolo/yolov8_pose/ccl_build/yolov8n_pose_backbone_OpSet16_optimized_asym_int8_q_shaped_QC_U_1d7_16MB_4kB_128GBps_128GBps_16_OFF_x1_x1/run/20260718_121525_1ac635


2026-07-18 12:16 - INFO - epu - chimera_job - Combined plots generated and saved to: 
/quadric/sdk-cli/examples/models/yolo/yolov8_pose/ccl_build/yolov8n_pose_backbone_OpSet16_optimized_asym_int8_q_shaped_QC_U_1d7_16MB_4kB_128GBps_128GBps_16_OFF_x1_x1/run/20260718_121525_1ac635/data/yolov8n_pose_backbone_OpSet16_optimized_asym_int8_q_shaped_QC_U_1d7_16MB_4kB_128GBps_128GBps_16_OFF_x1_x1.combined.png





'/quadric/sdk-cli/examples/models/yolo/yolov8_pose/ccl_build/yolov8n_pose_backbone_OpSet16_optimized_asym_int8_q_shaped_QC_U_1d7_16MB_4kB_128GBps_128GBps_16_OFF_x1_x1/run/20260718_121525_1ac635/data'

Postprocessing

At first, post-processing for inference results is done here to retrieve bounding boxes and keypoints.

import numpy as np
from PIL import Image

from sdk_cli.node_builtins.classical.yolo_postprocessing import (
    get_postprocess_handle,
    boxes_detections,
)
from sdk_cli.utils.models.yolo import YOLOModelVariant

postprocess_handle = get_postprocess_handle(YOLOModelVariant.YOLOv8_Pose_Estimation)

bboxes_per_inference_engine = {}
keypoints_per_inference_engine = {}
keypoints_scores_per_inference_engine = {}

for inference_engine, all_outputs in outputs_per_inference_engine.items():
    bboxes_per_inference_engine[inference_engine] = []
    keypoints_per_inference_engine[inference_engine] = []
    keypoints_scores_per_inference_engine[inference_engine] = []
    for outputs, image_path in zip(all_outputs, all_image_paths):
        detections, _, keypoints, keypoints_scores = boxes_detections(
            postprocess_handle,  # Handles for postprocess
            np.array(Image.open(image_path)),  # numpy image
            outputs,  # Outputs of the inference engine
            dataset_input_size,  # Model image size
        )
        bboxes_per_inference_engine[inference_engine].append(detections)
        keypoints_per_inference_engine[inference_engine].append(keypoints)
        keypoints_scores_per_inference_engine[inference_engine].append(keypoints_scores)

Display Bounding Boxes and Keypoints

import matplotlib.pyplot as plt
from sdk_cli.node_builtins.outputs.bbox_label_visualizer import draw_bbox
from sdk_cli.node_builtins.outputs.keypoints_visualizer import draw_keypoints

%matplotlib inline

pic_len = len(engines) + 1

for i in range(len(all_images)):
    ax, idx = {}, 1
    fig = plt.figure(figsize=(6 * pic_len, 6), tight_layout=True)

    ax[idx] = fig.add_subplot(int("1%s%s" % (pic_len, idx)))
    ax[idx].imshow(Image.open(all_image_paths[i]))
    ax[idx].set_title("Original Image")
    ax[idx].axis("off")

    for (inference_engine, all_bboxes), all_keypoints, all_scores in zip(
        bboxes_per_inference_engine.items(),
        keypoints_per_inference_engine.values(),
        keypoints_scores_per_inference_engine.values(),
    ):
        frame = np.array(Image.open(all_image_paths[i]))
        for keypoints, keypoints_scores in zip(all_keypoints[i], all_scores[i]):
            frame = draw_keypoints(frame, keypoints, keypoints_scores)
        frame = draw_bbox(frame, all_bboxes[i])
        idx += 1
        ax[idx] = fig.add_subplot(int("1%s%s" % (pic_len, idx)))
        ax[idx].imshow(frame)
        ax[idx].set_title(f"{MODEL_NAME}: {str(inference_engine)}")
        ax[idx].axis("off")

    fig.show()

Citation

@software{yolov8_ultralytics,
  author = {Glenn Jocher and Ayush Chaurasia and Jing Qiu},
  title = {Ultralytics YOLOv8},
  version = {8.0.0},
  year = {2023},
  url = {https://github.com/ultralytics/ultralytics},
  orcid = {0000-0001-5950-6979, 0000-0002-7603-6750, 0000-0003-3783-7069},
  license = {AGPL-3.0}
}

YOLOv8 Pose Estimation Environment Setup Model Selection Model Generation and ONNX Export Split ONNX Quantization Compilation Demo Inference as Batch Inference Postprocessing Display Bounding Boxes and Keypoints Citation

Model Demo: YOLOv8 Detection

Model Demo: YOLOP Detection and Segmentation

Table of Contents

Home

What's New

Chimera Software User Guide

Introduction to the Chimera SDK

Chimera SDK Quick Start Guide

Chimera SDK Command Line Interface (CLI)

CGC API Reference

Tutorial: Using SDK as a Library

Tutorials & Model Demos

Quantization Tutorials

Custom Op Tutorials

SDK Tutorials

Model Demos

Model Demo: Llama-2 15M (Baby Llama-2)

Model Demo: QWEN3 8B End-to-End CGC and ISS Execution

Model Demo: QWEN3 Prefill All Decoders

Model Demo: DeepSeek-R1-Distill-Qwen-1.5B End-to-End CGC and ISS Execution

Model Demo: QWEN3 Single Decoder

Model Demo: Qwen2.5-0.5B INT8 Quantization Pipeline

Model Demo: ConvNeXt Detection

Model Demo: QWEN3 Prefill Decoder Validation

Model Demo: ConvNeXt Segmentation

Model Demo: Classifiers Zoo

Model Demo: Detectors Zoo - MMDetection

Model Demo: Segmentors Zoo - MMSegmentation

Model Demo: Pose Estimators Zoo - MMPose

Model Demo: Detectors3D Zoo - MMDetection3D

MODEL Demo: Optical Character Recognition (OCR) Zoo - MMOCR

Model Demo: YOLOv3 Object Detection

Model Demo: YOLOv4 Object Detection

Model Demo: YOLOv5 Detection

Model Demo: YOLOv5 Detection and Segmentation

Model Demo: YOLOR Detection

Model Demo: YOLOX End-to-End Detection

Model Demo: YOLOv7 Detection

Model Demo: YOLOv8 Detection

Model Demo: YOLOv8 Pose Estimation

Model Demo: YOLOP Detection and Segmentation

Model Demo: QAT Vision Transformer (ViT)

Model Demo: QAT Swin Transformer

Model Demo: Mediapipe Face Pipeline

Demo: DOOM Renderer on Chimera GPNPU

Model Demo: Mediapipe Hand Pipeline

Model Demo: Whisper Tiny (Encoder + Decoder)

Model Demo: L2CS Fine-Grained Gaze Estimation

Model Demo: ASVspoof2021 LA Anti-Spoofing (LFCC-LCNN-BiLSTM)

Model Demo: UNET Tumor Segmentation

Model Demo: DETR Encoder

Model Demo: FFNet Segmentation

Model Demo: Centernet Detection

Model Demo: RetinaNet End-to-End Detection

Model Demo: Blazepose Pose Estimation

Model Demo: Pose Resnet Human Pose Estimation

Model Demo: MaskRCNN Detection and Segmentation

Model Demo: Keypoint R-CNN

Model Demo: Faster R-CNN Detection

Model Demo: FCOS Detection

Model Demo: DDRNet Classificationls

Model Demo: PI0.5 End-to-End VLA Inference

Model Demo: BEVFormer End-to-End 3D Detection

Multicore Demo

Chimera Graph Compiler (CGC)

Chimera Compute Library (CCL) API

Chimera LLVM C++ Compiler

Chimera Instruction Set Simulator (ISS)

Debugging

Chimera SDK Licensing Policy Documentation

Glossary

Datasheet - Chimera QC Series GPNPU Processors

All systems operational

270 East Lane
Burlingame, CA 94010
USA

1-844-GPNPU-00
(1-844-476-7800)

PRODUCTS

COMPANY

NEWS + INSIGHTS

This documentation is preliminary and confidential. It is subject to change. Quadric does not give any warranty express or implied that the contents will be complete or accurate or up to date. The company shall not be liable for any loss, actions, claims, proceedings, demands or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

UH-OH

It looks like you don’t have access to that feature yet

UH-OH

Model Demo: YOLOv8 Pose Estimation

YOLOv8 Pose Estimation

Environment Setup

Model Selection

Model Generation and ONNX Export

Split ONNX

Quantization

Compilation

Demo

Inference as Batch

Inference

Run Statistics for PoseResnet Backbone

Postprocessing

Display Bounding Boxes and Keypoints

Citation

PRODUCTS

COMPANY

NEWS + INSIGHTS

Sign in to your account