- Object Detection

- Object Detection

All of the original and compressed models can be downloaded easily on the Model Compressor Model Zoo.

You can get Compressed results with Automatic Compression and Compressed (Adv.) results with Advanced Compression.


PyTorch

ModelBest PracticeTypeDatasetmAP(0.5) (%)mAP(0.5:0.95)(%)FLOPs (M)Params (M)Latency (ms)Model Size (MB)
YOLOXOriginalCOCO68.049.7156006.2054.2112239.46207.37
YOLOXGoogle ColabCompressed-1COCO67.16 (-0.84)48.64 (-1.06)101804.06 (1.53x)19.96 (2.7x)8502.72 (1.44x)76.61 (2.7x)
YOLOXGoogle ColabCompressed-2COCO61.43 (-6.57)43.23 (-5.47)38607.03 (4.04x)4.93 (11.0x)4235.37 (2.89x)19.17 (10.80x)
  • The model’s latency is measured on Raspberry Pi 4B (1.5GHz ARM Cortex).
  • Options: FP32, ONNX runtime

TensorFlow-Keras

ModelBest PracticeTypeDatasetmAP(0.5) (%)mAP(0.5:0.95)(%)FLOPs (M)Params (M)Latency (ms)Model Size (MB)
YOLOv4OriginalPASCAL VOC82.22-61871.8265.3264318.70262.90
YOLOv4Google ColabCompressed-1PASCAL VOC87.23 (+5.01)-11459.69 (5.4x)10.59 (6.17x)28651.70 (2.16x)44.12 (5.96x)
YOLOv4Google ColabCompressed-2PASCAL VOC87.91 (+5.69)-14442.96 (4.28x)10.71 (6.1x)28976.40 (2.14x)44.36 (5.93x)
  • YOLOv4 model with EfficientNet B1 based backbone.
  • The model’s latency is measured on Raspberry Pi 4B (1.5GHz ARM Cortex).
  • Options: FP32, TFLite