Skip to content
Permalink
master
Switch branches/tags
Go to file
PiperOrigin-RevId: 372579796
6 contributors

Users who have contributed to this file

@tombstone @a-googler @vighneshbirodkar @srjoglekar246 @pkulzc @khanhlvg

TensorFlow 2 Detection Model Zoo

TensorFlow 2.2 Python 3.6

We provide a collection of detection models pre-trained on the COCO 2017 dataset. These models can be useful for out-of-the-box inference if you are interested in categories already in those datasets. You can try it in our inference colab

They are also useful for initializing your models when training on novel datasets. You can try this out on our few-shot training colab.

Please look at this guide for mobile inference.

Finally, if you would like to train these models from scratch, you can find the model configs in this directory (also in the linked tar.gzs).

Model name Speed (ms) COCO mAP Outputs
CenterNet HourGlass104 512x512 70 41.9 Boxes
CenterNet HourGlass104 Keypoints 512x512 76 40.0/61.4 Boxes/Keypoints
CenterNet HourGlass104 1024x1024 197 44.5 Boxes
CenterNet HourGlass104 Keypoints 1024x1024 211 42.8/64.5 Boxes/Keypoints
CenterNet Resnet50 V1 FPN 512x512 27 31.2 Boxes
CenterNet Resnet50 V1 FPN Keypoints 512x512 30 29.3/50.7 Boxes/Keypoints
CenterNet Resnet101 V1 FPN 512x512 34 34.2 Boxes
CenterNet Resnet50 V2 512x512 27 29.5 Boxes
CenterNet Resnet50 V2 Keypoints 512x512 30 27.6/48.2 Boxes/Keypoints
CenterNet MobileNetV2 FPN 512x512 6 23.4 Boxes
CenterNet MobileNetV2 FPN Keypoints 512x512 6 41.7 Keypoints
EfficientDet D0 512x512 39 33.6 Boxes
EfficientDet D1 640x640 54 38.4 Boxes
EfficientDet D2 768x768 67 41.8 Boxes
EfficientDet D3 896x896 95 45.4 Boxes
EfficientDet D4 1024x1024 133 48.5 Boxes
EfficientDet D5 1280x1280 222 49.7 Boxes
EfficientDet D6 1280x1280 268 50.5 Boxes
EfficientDet D7 1536x1536 325 51.2 Boxes
SSD MobileNet v2 320x320 19 20.2 Boxes
SSD MobileNet V1 FPN 640x640 48 29.1 Boxes
SSD MobileNet V2 FPNLite 320x320 22 22.2 Boxes
SSD MobileNet V2 FPNLite 640x640 39 28.2 Boxes
SSD ResNet50 V1 FPN 640x640 (RetinaNet50) 46 34.3 Boxes
SSD ResNet50 V1 FPN 1024x1024 (RetinaNet50) 87 38.3 Boxes
SSD ResNet101 V1 FPN 640x640 (RetinaNet101) 57 35.6 Boxes
SSD ResNet101 V1 FPN 1024x1024 (RetinaNet101) 104 39.5 Boxes
SSD ResNet152 V1 FPN 640x640 (RetinaNet152) 80 35.4 Boxes
SSD ResNet152 V1 FPN 1024x1024 (RetinaNet152) 111 39.6 Boxes
Faster R-CNN ResNet50 V1 640x640 53 29.3 Boxes
Faster R-CNN ResNet50 V1 1024x1024 65 31.0 Boxes
Faster R-CNN ResNet50 V1 800x1333 65 31.6 Boxes
Faster R-CNN ResNet101 V1 640x640 55 31.8 Boxes
Faster R-CNN ResNet101 V1 1024x1024 72 37.1 Boxes
Faster R-CNN ResNet101 V1 800x1333 77 36.6 Boxes
Faster R-CNN ResNet152 V1 640x640 64 32.4 Boxes
Faster R-CNN ResNet152 V1 1024x1024 85 37.6 Boxes
Faster R-CNN ResNet152 V1 800x1333 101 37.4 Boxes
Faster R-CNN Inception ResNet V2 640x640 206 37.7 Boxes
Faster R-CNN Inception ResNet V2 1024x1024 236 38.7 Boxes
Mask R-CNN Inception ResNet V2 1024x1024 301 39.0/34.6 Boxes/Masks
ExtremeNet (deprecated) -- -- Boxes
ExtremeNet -- -- Boxes