| MobileNetV2-Asymmetric | 4,666 FPS @4MB, 128GB/s external AXI bandwidth, 8MACs per PE, 1.7GHz |
| ResNet-50 | 1,330 FPS |
| ATSS Backbone (ResNet101) | |
| ATSS Backbone (ResNet50) | |
| AutoAssign Backbone (ResNet50) | |
| BlazePose Detector | |
| BoxInst Backbone+Neck (ResNet101 800x800) | |
| BoxInst Backbone+Neck (ResNet50 800x800) | |
| CenterNet | |
| CenterNet Update Backbone (ResNet50) | |
| CondInst Backbone+Neck (ResNet50 800x800) | |
| ConvNeXt Small for UPerNet (640x640) | |
| ConvNeXt Small for UPerNet (928x1600) | |
| ConvNeXt Tiny for Mask RCNN (800x800) | |
| ConvNeXt Tiny for Mask RCNN with Neck (800x800) | |
| ConvNeXt Tiny for UPerNet (640x640) | |
| ConvNeXt Tiny for UPerNet (928x1600) | |
| CS3-DarkNet + Focus Layer Large | |
| CS3-DarkNet + Focus Layer Medium | |
| CS3-DarkNet Large | |
| CS3-DarkNet Medium | |
| CSPDarkNet-53 | |
| CSPDarkNet-53 (256x256) | |
| Darknet-53 | |
| DBNet Backbone+Neck (ResNet18+FPNC) for ICDAR2015 | |
| DBNet Backbone+Neck (ResNet18+FPNC) for Total-Text | |
| DBNet Backbone+Neck (ResNet50+FPNC) for ICDAR2015 | |
| DBNetpp Backbone (ResNet50+FPNC) for ICDAR2015 | |
| DDRNet_23_slim Classifier | |
| Decoupled Light SOLO Backbone (ResNet50) | |
| Decoupled SOLO Backbone (ResNet50) | |
| DeepLabV3 Backbone (MobileNetV2 512x512) | |
| DeepLabV3+ Backbone (MobileNetV2 512x512) | |
| DeepLabV3 Backbone (U-Net, 256x256) for HRF | |
| DeepLabV3 (U-Net, 128x128) for CHASEDB1 | |
| DeepLabV3 (U-Net, 128x128) for STARE | |
| DeepLabV3 (U-Net, 128x128) for STARE with Dice Metrics | |
| DeepLabV3 (U-Net, 64x64) for DRIVE | |
| DeepLabV3 (U-Net, 64x64) for DRIVE with Dice Metrics | |
| DeepLabV3 with MobilenetV2 Backbone | |
| DETR Segmenter Backbone | |
| Dynamic R-CNN Backbone+Neck (ResNet50 800x800) | |
| EdgeTPU-optimized EfficientNet Large | |
| EdgeTPU-optimized EfficientNet Medium | |
| EdgeTPU-optimized EfficientNet Pruned Large | |
| EdgeTPU-optimized EfficientNet Pruned Small | |
| EdgeTPU-optimized EfficientNet Small | |
| EfficientNet-Lite0 | |
| Faster R-CNN Backbone | |
| Faster R-CNN Backbone+Neck (RegNetX-0.4GF+FPN, multi-scale training, 800x800) | |
| Faster R-CNN Backbone+Neck (RegNetX-0.8GF+FPN, multi-scale training, 800x800) | |
| Faster R-CNN Backbone+Neck (RegNetX-1.6GF+FPN, multi-scale training, 800x800) | |
| Faster R-CNN Backbone+Neck (RegNetX-3.2GF+FPN, 800x800) | |
| Faster R-CNN Backbone+Neck (RegNetX-3.2GF+FPN, multi-scale training, 800x800) | |
| Faster R-CNN Backbone+Neck (ResNet101+FPN, 800x800) | |
| Faster R-CNN Backbone+Neck (ResNet50+FPN, 800x800) | |
| Faster R-CNN Backbone+Neck (ResNet50+PAFPN, 800x800) | |
| FBNet-C-100 | |
| FCENet Backbone (ResNet50+FPN) for ICDAR2015 | |
| FCENet (ResNet50+FPN) for Total-Text | |
| FCN (HRNetV2p-W18, 512x1024) | |
| FCN (HRNetV2p-W18-Small 512x1024) | |
| FCN (HRNetV2p-W18-Small, 512x512) for ADE20K | |
| FCN (HRNetV2p-W48, 512x1024) | |
| FCN (MobileNetV2 512x1024) | |
| FCN (MobileNetV2, 512x512) for ADE20K | |
| FCN (U-Net, 128x128) for CHASEDB | |
| FCN (U-Net, 128x128) for CHASEDB with Dice Metrics | |
| FCN (U-Net, 128x128) for STARE | |
| FCN (U-Net, 256x256) for HRF | |
| FCN (U-Net, 64x64) for DRIVE | |
| FCN (U-Net, 64x64) for DRIVE with Dice Metrics | |
| FCOS | |
| FCOS Backbone+Neck (ResNet101, 800x800) by Caffe | |
| FCOS Backbone+Neck (ResNet50, 800x800) by Caffe | |
| FFNet-150S | |
| FFNet-40S | |
| FFNet-54S | |
| FFNet-78S | |
| FFNet-86S | |
| FOVEA (ResNet101) | |
| FOVEA (ResNet50) | |
| FreeAnchor for 3D Object Detection Predictor (RegNet-0.4GF+FPN, Strong Augmentation) for nuScenes | |
| FreeAnchor for 3D Object Detection Predictor (RegNet-1.6GF+FPN) for nuScenes | |
| FreeAnchor for 3D Object Detection Predictor (RegNet-1.6GF+FPN, Strong Augmentation) for nuScenes | |
| FreeAnchor for 3D Object Detection Predictor (RegNet-3.2GF+FPN, Strong Augmentation) for nuScenes | |
| FreeAnchor for 3D Object Detection Predictor (RegNet-3.2GFL+FPN) for nuScenes | |
| FreeAnchor (ResNet101) | |
| FreeAnchor (ResNet50) | |
| FSAF (ResNet50) | |
| GFL Backbone (ResNet101) | |
| GFL Backbone (ResNet50) | |
| Gluon ResNet101 | |
| Gluon ResNeXt101 | |
| Keypoint R-CNN Backbone | |
| L2CS-Net | |
| LAD Backbone (Student:PAA ResNet101, Teacher:PAA ResNet50) | |
| LAD Backbone (Student: PAA ResNet50, Teacher: PAA ResNet101) | |
| LD (Student: ResNet18, Teacher: ResNet101, 800x800) | |
| LD (Student: ResNet34, Teacher: ResNet101, 800x800) | |
| LD (Student: ResNet50, Teacher: ResNet101, 800x800) | |
| Llama-2-15M | |
| Mask R-CNN Backbone | |
| Mask R-CNN Backbone+Neck (RegNetX-0.4GF+FPN, multi-scale training, poly LR schedule, 800x800) | |
| Mask R-CNN Backbone+Neck (RegNetX-0.8GF+FPN, multi-scale training, poly LR schedule, 800x800) | |
| Mask R-CNN Backbone+Neck (RegNetX-1.6GF+FPN, multi-scale training, poly LR schedule, 800x800) | |
| Mask R-CNN Backbone+Neck (RegNetX-3.2GF+FPN, 800x800) | |
| Mask R-CNN Backbone+Neck (RegNetX-3.2GF+FPN, multi-scale training, 800x800) | |
| Mask R-CNN Backbone+Neck (ResNet101+FPN, 800x800) | |
| Mask R-CNN Backbone+Neck (ResNet50+FPN, 800x800) | |
| Mask R-CNN Keypoint Detection | |
| Mediapipe Face Detection Full Range | |
| Mediapipe Face Detection Short Range | |
| Mediapipe Face Landmark | |
| Mediapipe Hand Landmark Full | |
| Mediapipe Hand Landmark Lite | |
| Mediapipe Iris Landmark | |
| Mediapipe Object Detection 3d Chair 1stage | |
| Mediapipe Object Detection 3d Sneakers 1stage | |
| Mediapipe Palm Detection Full | |
| Mediapipe Palm Detection Lite | |
| Mediapipe SSDLITE Object Detection | |
| MnasNet-0.75 | |
| MnasNet-1.0 | |
| MnasNet-100 | |
| MobileNetV2-110D | |
| MobileNetV2-140 | |
| MobileNetV2-Symmetric | |
| OCRNet Backbone (HRNetV2p-W18 512x1024) | |
| OCRNet Backbone (HRNetV2p-W18, 512x512) for ADE20K | |
| OCRNet Backbone (HRNetV2p-W18-Small 512x1024) | |
| PAA Backbone (ResNet101) | |
| PAA Backbone (ResNet50) | |
| PANet Backbone (ResNet18+FPEM_FFM) for CTW1500 | |
| PANet Backbone (ResNet18+FPEM_FFM) for ICDAR2015 | |
| PGD Backbone (ResNet101) for KITTI | |
| PISA (RetinaNet ResNet50) | |
| PointPillars Predictor (SECOND+FPN) for nuScenes | |
| Pose ResNet Backbone | |
| PSENet Backbone (ResNet50+FPNF) for CTW1500 | |
| PSPNet Backbone (MobileNetV2 512x512) | |
| PSPNet Backbone (U-Net, 128x128) for CHASEDB | |
| Res2Net50 (26w, 4s) | |
| Res2Net-50 (48w, 2s) | |
| Res2NeXt-50 | |
| Resnet-101 | |
| ResNet101-C | |
| ResNet-101 (LinearClsHead) | |
| ResNet101-S | |
| ResNet-152 | |
| ResNet152-C | |
| ResNet-152 (LinearClsHead) | |
| ResNet152-S | |
| ResNet-18 | |
| ResNet-18 (LinearClsHead) | |
| ResNet26 | |
| ResNet32-TS | |
| ResNet33-TS | |
| ResNet-34 | |
| ResNet-34 (LinearClsHead) | |
| ResNet-50 (Barlow Twins) | |
| ResNet50 (BlurPool) | |
| ResNet-50 (BYOL) | |
| ResNet50-C | |
| ResNet-50 (DenseCL) | |
| ResNet50 (GroupNorm) | |
| ResNet-50 (LinearClsHead) | |
| ResNet-50 (MoCoV2) | |
| ResNet-50 (MoCoV3) | |
| ResNet50-S | |
| ResNet-50 (SimCLR) | |
| ResNet-50 (SimSiam) | |
| ResNet-50 (SparK) | |
| ResNet-50 (SwAV) | |
| ResNeXt101 (32x4d) | |
| ResNeXt101 (32x8d) | |
| ResNeXt101 (64x4d) | |
| ResNeXt26-TS | |
| ResNeXt50 (32x4d) | |
| RetinaNet | |
| RetinaNet Backbone+Neck (RegNetX-0.8GF+FPN, 800x800) | |
| RetinaNet Backbone+Neck (RegNetX-1.6GF+FPN, 800x800) | |
| RetinaNet Backbone+Neck (RegNetX-3.2GF+FPN, 800x800) | |
| RetinaNet Backbone+Neck (ResNet50, 800x800) for Open Images | |
| RetinaNet for GHM (ResNet101, 800x800) | |
| RetinaNet for GHM (ResNet50, 800x800) | |
| RetinaNet for ResNet Strikes Back (ResNet50, 800x800) | |
| RetinaNet (ResNet101) | |
| RetinaNet (ResNet18) | |
| RetinaNet (ResNet50 800x640) | |
| RIFormer S12 | |
| RIFormer S24 | |
| RIFormer S24 (384x384) | |
| RPN (ResNet101, 800x800) | |
| RPN (ResNet50, 800x800) | |
| SelecSLS-42B | |
| SelecSLS-60 | |
| SelecSLS-60B | |
| Semantic FPN Backbone+Neck (ResNet50 512x1024) | |
| Semantic FPN Backbone+Neck (ResNetV1c50+FPN, 512x512) for ADE20K | |
| Semantic FPN (ResNet101, 512x1024) | |
| Semantic FPN (ResNet101, 512x512) for ADE20K | |
| Semantic FPN (ResNet50, 512x1024) | |
| Semantic FPN (ResNet50, 512x512) for ADE20K | |
| SMOKE Backbone (DLANet34 GroupNorm+DLANeck) for KITTI | |
| SOLO Backbone (ResNet50) | |
| SPNASNet-100 | |
| SSD-Lite (MobileNetV2 320x320) | |
| SSL ResNet-18 | |
| SSL ResNet-50 | |
| SWSL ResNet-18 | |
| SWSL ResNet-50 | |
| Textsnake Backbone (ResNet50+FPN_Unet) for CTW1500 | |
| TF-trained EfficientNet-Lite0 | |
| TF-trained EfficientNet-Lite1 | |
| TF-trained EfficientNet-Lite2 | |
| TF-trained EfficientNet-Lite3 | |
| TF-trained EfficientNet-Lite4 | |
| Top-down heatmap-based Animal Keypoint Detection Backbone (ResNet101 256x256) | |
| Top-down heatmap-based Animal Keypoint Detection Backbone (ResNet152 256x256) | |
| Top-down heatmap-based Animal Keypoint Detection Backbone (ResNet50 256x256) | |
| Top-down heatmap-based Face Landmark Detection Backbone (MobileNetV2 256x256) | |
| Top-down heatmap-based Face Landmark Detection Backbone (ResNet50 256x256) | |
| Top-down heatmap-based Face Landmark Detection (Hourglass52 256x256) | |
| Top-down heatmap-based Hand Pose Estimation Backbone (MobileNetV2 256x256) | |
| Top-down heatmap-based Hand Pose Estimation Backbone (ResNet50 256x256) | |
| Top-down heatmap-based Hand Pose Estimation (Hourglass52 256x256) | |
| Top-down heatmap-based Human Whole-Body Pose Estimation Backbone (ResNet101 256x192) | |
| Top-down heatmap-based Human Whole-Body Pose Estimation Backbone (ResNet101 384x288) | |
| Top-down heatmap-based Human Whole-Body Pose Estimation Backbone (ResNet152 256x192) | |
| Top-down heatmap-based Human Whole-Body Pose Estimation Backbone (ResNet152 384x288) | |
| Top-down heatmap-based Human Whole-Body Pose Estimation Backbone (ResNet50 256x192) | |
| Top-down heatmap-based Human Whole-Body Pose Estimation Backbone (ResNet50 384x288) | |
| Top-down heatmap-based pose estimation Backbone (MobileNetV2 256x192) | |
| Top-down heatmap-based pose estimation Backbone (ResNet101, 256x192) for COCO | |
| Top-down heatmap-based pose estimation Backbone (ResNet101, 256x192) for CrowdPose | |
| Top-down heatmap-based pose estimation Backbone (ResNet101, 384x288) for COCO | |
| Top-down heatmap-based pose estimation Backbone (ResNet152, 256x192) for CrowdPose | |
| Top-down heatmap-based pose estimation Backbone (ResNet50, 256x192) for COCO | |
| Top-down heatmap-based pose estimation Backbone (ResNet50, 256x192) for CrowdPose | |
| Top-down heatmap-based pose estimation (Hourglass, 256x256) for COCO | |
| Top-down heatmap-based pose estimation (Hourglass, 256x256) for MPII | |
| Top-down heatmap-based pose estimation (Hourglass, 384x384) for MPII | |
| Top-down heatmap-based pose estimation (Hourglass52 384x384) | |
| Top-down heatmap-based pose estimation (MobileNetV2, 384x288) for COCO | |
| Top-down heatmap-based pose estimation (MSPN50, 256x192) | |
| Top-down heatmap-based pose estimation (RSN18, 256x192) | |
| Top-down heatmap-based pose estimation (RSN50, 256x192) | |
| Top-down heatmap-based pose estimation (VGG16 256x192) | |
| Top-down regression-based pose estimation Backbone (MobileNetV2, 256x192) for COCO | |
| Top-down regression-based pose estimation Backbone (ResNet101, 256x192) for COCO | |
| Top-down regression-based pose estimation Backbone (ResNet101, 256x256) for MPII | |
| Top-down regression-based pose estimation Backbone (ResNet152, 256x192) for COCO | |
| Top-down regression-based pose estimation Backbone (ResNet152, 256x256) for MPII | |
| Top-down regression-based pose estimation Backbone (ResNet152, 384x288) for COCO | |
| Top-down regression-based pose estimation Backbone (ResNet50, 256x192) for COCO | |
| Top-down regression-based pose estimation Backbone (ResNet50, 256x256) for MPII | |
| UltraFace_WIDE | |
| U-Net (Brain Tumor) | |
| UPerNet Backbone (ResNet50 512x1024) | |
| VGG11 | |
| VGG11 (BatchNorm) | |
| VGG11 (BatchNorm, ClsHead) | |
| VGG11 (ClsHead) | |
| VGG13 | |
| VGG13 (BatchNorm) | |
| VGG13 (BatchNorm, ClsHead) | |
| VGG13 (ClsHead) | |
| VGG16 | |
| VGG16 (BatchNorm) | |
| VGG16 (BatchNorm, ClsHead) | |
| VGG16 (ClsHead) | |
| VGG19 | |
| VGG19 (BatchNorm) | |
| VGG19 (BatchNorm, ClsHead) | |
| VGG19 (ClsHead) | |
| ViT (Vision Transformer) | |
| Wide ResNet-101 | |
| Wide ResNet-50 | |
| YOLOF Backbone (ResNet50) | |
| YOLOP Backbone | |
| YOLOR_P6 Backbone | |
| YOLOv3 | |
| YOLOv3 (Darknet53, 320x320) | |
| YOLOv3 (Darknet53, 416x416) | |
| YOLOv3 (Darknet53, 608x608) | |
| YOLOv3 (MobileNetv2, 320x320) | |
| YOLOv3 (MobileNetV2 416x416) | |
| YOLOv4 | |
| YOLOv5L Backbone | |
| YOLOv5l Segmentation Backbone | |
| YOLOv5M Backbone | |
| YOLOv5m Segmentation Backbone | |
| YOLOv5N6 Backbone | |
| YOLOv5N Backbone | |
| YOLOv5n Segmentation Backbone | |
| YOLOv5S Backbone | |
| YOLOv5s Segmentation | |
| YOLOv5s Segmentation Backbone | |
| YOLOv7 | |
| YOLOv7-Tiny Backbone | |
| YOLOv8N Backbone | |
| YOLOv8n Pose Estimation Backbone | |
| YOLOv8S Backbone | |
| YOLOv8s Pose Estimation Backbone | |
| YOLOX-M (No Concat) | |
| YOLOX-S (640x640) | |
| YOLOX-S (No Concat) | |
| YOLOX-Tiny (416x416) | |
| YOLOX-Tiny (No Concat) | |
'%3e%3crect%20width='512'%20height='512'%20rx='60'%3e%3c/rect%3e%3cpath%20fill='%23fff'%20d='M355.904%20100H408.832L293.2%20232.16L429.232%20412H322.72L239.296%20302.928L143.84%20412H90.8805L214.56%20270.64L84.0645%20100H193.28L268.688%20199.696L355.904%20100ZM337.328%20380.32H366.656L177.344%20130.016H145.872L337.328%20380.32Z'%3e%3c/path%3e%3c/g%3e%3cdefs%3e%3cclipPath%20id='clip0_84_15698'%3e%3crect%20width='512'%20height='512'%20fill='%23fff'%3e%3c/rect%3e%3c/clipPath%3e%3c/defs%3e%3c/svg%3e)
