- Python 3.7
- PaddlePaddle Daily version
- CUDA 10.1
- cuDNN 7.5
- NCCL 2.4.8
- All models were trained and tested in the COCO17 dataset.
- Unless special instructions, all the ResNet backbone network using ResNet-B structure.
- Inference time (FPS): The reasoning time was calculated on a Tesla V100 GPU by
tools/eval.pytesting all validation sets in FPS (number of pictures/second). CuDNN version is 7.5, including data loading, network forward execution and post-processing, and Batch size is 1.
- We adopt and Detectron in the same training strategy.
- 1x strategy indicates that when the total batch size is 8, the initial learning rate is 0.01, and the learning rate decreases by 10 times after 8 epoch and 11 epoch, respectively, and the final training is 12 epoch.
- 2X strategy is twice as much as strategy 1X, and the learning rate adjustment position is twice as much as strategy 1X.
Paddle provides a skeleton network pretraining model based on ImageNet. All pre-training models were trained by standard Imagenet 1K dataset. Res Net and Mobile Net are high-precision pre-training models obtained by cosine learning rate adjustment strategy or SSLD knowledge distillation training. Model details are available at PaddleClas.
Please refer toFaster R-CNN
Please refer toMask R-CNN
Please refer toCascade R-CNN
Please refer toYOLOv3
Please refer toSSD
Please refer toFCOS
Please refer toSOLOv2
Please refer toPP-YOLO
请参考TTFNet
Please refer toGroup Normalization
Please refer toDeformable ConvNets v2
Please refer toHRNets
Please refer toRes2Net
Please refer toGFL
Please refer toPicoDet
Please refer toPP-YOLOE
Please refer toYOLOX
Please refer toYOLOv5
Please refer toYOLOv6
Please refer toYOLOv7
Please refer toS2ANet
Please refer to PP-TinyPose
Please refer to HRNet
Please refer to HigherHRNet
Please refer to DeepSORT
Please refer to JDE
Please refer to FairMOT
Please refer to ByteTrack