PytorchInsight

This is a pytorch lib with state-of-the-art architectures, pretrained models and real-time updated results.

This repository aims to accelarate the advance of Deep Learning Research, make reproducible results and easier for doing researches, and in Pytorch.

Including Papers (to be updated):

Attention Models

SENet: Squeeze-and-excitation Networks _(paper)

SKNet: Selective Kernel Networks _(paper)

CBAM: Convolutional Block Attention Module _(paper)

GCNet: GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond _(paper)

BAM: Bottleneck Attention Module _(paper)

SGENet: Spatial Group-wise Enhance: Enhancing Semantic Feature Learning in Convolutional Networks _(paper)

SRMNet: SRM: A Style-based Recalibration Module for Convolutional Neural Networks _(paper)

Non-Attention Models

OctNet: Drop an Octave: Reducing Spatial Redundancy in Convolutional Neural Networks with Octave Convolution _(paper)

imagenet_tricks.py: Bag of Tricks for Image Classification with Convolutional Neural Networks _(paper)

Trained Models and Performance Table

Single crop validation error on ImageNet-1k (center 224x224 crop from resized image with shorter side = 256).

	classifiaction training settings for media and large models
Details	RandomResizedCrop, RandomHorizontalFlip; 0.1 init lr, total 100 epochs, decay at every 30 epochs; SGD with naive softmax cross entropy loss, 1e-4 weight decay, 0.9 momentum, 8 gpus, 32 images per gpu
Examples	ResNet50

	classifiaction training settings for mobile/small models
Details	RandomResizedCrop, RandomHorizontalFlip; 0.4 init lr, total 300 epochs, 5 linear warm up epochs, cosine lr decay; SGD with softmax cross entropy loss and label smoothing 0.1, 4e-5 weight decay on conv weights, 0 weight decay on all other weights, 0.9 momentum, 8 gpus, 128 images per gpu
Examples	ShuffleNetV2

Typical Training & Testing Tips:

You may need to add some models to use in classification/models/imagenet/__init__.py, e.g., add

from .shufflenetv2 import *

Small Models

ShuffleNetV2_1x

python -m torch.distributed.launch --nproc_per_node=8 imagenet_mobile.py --cos -a shufflenetv2_1x --data /path/to/imagenet1k/ \
--epochs 300 --wd 4e-5 --gamma 0.1 -c checkpoints/imagenet/shufflenetv2_1x --train-batch 128 --opt-level O0 # Triaing

python -m torch.distributed.launch --nproc_per_node=2 imagenet_mobile.py -a shufflenetv2_1x --data /path/to/imagenet1k/ \
-e --resume ../pretrain/shufflenetv2_1x.pth.tar --test-batch 100 --opt-level O0 # Testing, ~69.6% top-1 Acc

Large Models

SGE-ResNet

python -W ignore imagenet.py -a sge_resnet101 --data /path/to/imagenet1k/ --epochs 100 --schedule 30 60 90 \
--gamma 0.1 -c checkpoints/imagenet/sge_resnet101 --gpu-id 0,1,2,3,4,5,6,7 # Training

python -m torch.distributed.launch --nproc_per_node=8 imagenet_fast.py -a sge_resnet101 --data /path/to/imagenet1k/ \ 
--epochs 100 --schedule 30 60 90 --wd 1e-4 --gamma 0.1 -c checkpoints/imagenet/sge_resnet101 --train-batch 32 \ 
--opt-level O0 --wd-all --label-smoothing 0. --warmup 0 # Training (faster)

python -W ignore imagenet.py -a sge_resnet101 --data /path/to/imagenet1k/ --gpu-id 0,1 -e --resume ../pretrain/sge_resnet101.pth.tar \
# Testing ~78.8% top-1 Acc

python -m torch.distributed.launch --nproc_per_node=2 imagenet_fast.py -a sge_resnet101 --data /path/to/imagenet1k/ -e --resume \
../pretrain/sge_resnet101.pth.tar --test-batch 100 --opt-level O0 # Testing (faster) ~78.8% top-1 Acc

Classification

Model	#P	GFLOPs	Top-1 Acc	Top-5 Acc	Download	log
ShuffleNetV2_1x	2.28M	0.151	69.6420	88.7200	GoogleDrive	shufflenetv2_1x.log
ResNet50	25.56M	4.122	76.3840	92.9080	BaiduDrive(zuvx)	old_resnet50.log
Oct-ResNet50 (0.125)
SRM-ResNet50
SE-ResNet50	28.09M	4.130	77.1840	93.6720
SK-ResNet50	26.15M	4.185	77.5380	93.7000	BaiduDrive(tfwn)	sk_resnet50.log
BAM-ResNet50	25.92M	4.205	76.8980	93.4020	BaiduDrive(z0h3)	bam_resnet50.log
CBAM-ResNet50	28.09M	4.139	77.6260	93.6600	BaiduDrive(bram)	cbam_resnet50.log
GC-ResNet50
SGE-ResNet50	25.56M	4.127	77.5840	93.6640	BaiduDrive(gxo9)	sge_resnet50.log
ResNet101	44.55M	7.849	78.2000	93.9060	BaiduDrive(js5t)	old_resnet101.log
Oct-ResNet101 (0.125)
SRM-ResNet101
SE-ResNet101	49.33M	7.863	78.4680	94.1020	BaiduDrive(j2ox)	se_resnet101.log
SK-ResNet101	45.68M	7.978	78.7920	94.2680	BaiduDrive(boii)	sk_resnet101.log
BAM-ResNet101	44.91M	7.933	78.2180	94.0180	BaiduDrive(4bw6)	bam_resnet101.log
CBAM-ResNet101	49.33M	7.879	78.3540	94.0640	BaiduDrive(syj3)	cbam_resnet101.log
GC-ResNet101
SGE-ResNet101	44.55M	7.858	78.7980	94.3680	BaiduDrive(wqn6)	sge_resnet101.log

Detection

Model	#p	GFLOPs	Detector	Neck	AP50:95 (%)	AP50 (%)	AP75 (%)	Download
ResNet50	23.51M	88.032	Faster RCNN	FPN	37.5	59.1	40.6	BaiduDrive()
SGE-ResNet50	23.51M	88.149	Faster RCNN	FPN	38.7	60.8	41.7	BaiduDrive()
ResNet50	23.51M	88.032	Mask RCNN	FPN	38.6	60.0	41.9	BaiduDrive()
SGE-ResNet50	23.51M	88.149	Mask RCNN	FPN	39.6	61.5	42.9	BaiduDrive()
ResNet50	23.51M	88.032	Cascade RCNN	FPN	41.1	59.3	44.8	BaiduDrive()
SGE-ResNet50	23.51M	88.149	Cascade RCNN	FPN	42.6	61.4	46.2	BaiduDrive()
ResNet101	42.50M	167.908	Faster RCNN	FPN	39.4	60.7	43.0	BaiduDrive()
SGE-ResNet101	42.50M	168.099	Faster RCNN	FPN	41.0	63.0	44.3	BaiduDrive()
ResNet101	42.50M	167.908	Mask RCNN	FPN	40.4	61.6	44.2	BaiduDrive()
SGE-ResNet101	42.50M	168.099	Mask RCNN	FPN	42.1	63.7	46.1	BaiduDrive()
ResNet101	42.50M	167.908	Cascade RCNN	FPN	42.6	60.9	46.4	BaiduDrive()
SGE-ResNet101	42.50M	168.099	Cascade RCNN	FPN	44.4	63.2	48.4	BaiduDrive()

Model	#p	GFLOPs	Detector	Neck	AP small (%)	AP media (%)	AP large (%)	Download
ResNet50	23.51M	88.032	RetinaNet	FPN	19.9	39.6	48.3	BaiduDrive()
SE-ResNet50	26.04M	88.152	RetinaNet	FPN	20.7	41.3	50.0	BaiduDrive()
SK-ResNet50	24.11M	89.414	RetinaNet	FPN	20.2	40.9	50.4	BaiduDrive()
BAM-ResNet50	23.87M	89.804	RetinaNet	FPN	19.6	40.1	49.9	BaiduDrive()
CBAM-ResNet50	26.04M	88.302	RetinaNet	FPN	21.8	40.8	49.5	BaiduDrive()
SGE-ResNet50	23.51M	88.149	RetinaNet	FPN	21.8	41.2	49.9	BaiduDrive()

Citation

If you use related works in your research, please cite the paper:

@inproceedings{li2019selective,
  title={Selective Kernel Networks},
  author={Li, Xiang and Wang, Wenhai and Hu, Xiaolin and Yang, Jian},
  journal={IEEE Conference on Computer Vision and Pattern Recognition},
  year={2019}
}

@inproceedings{li2019spatial,
  title={Spatial Group-wise Enhance: Enhancing Semantic Feature Learning in Convolutional Networks},
  author={Li, Xiang and Hu, Xiaolin and Yang, Jian},
  journal={Arxiv},
  year={2019}
}

yangkang779 / PytorchInsight