fast-reid/projects/DistillReID

Distillation ReID

This project provides a training script of small model for both fast inference and high accuracy.

Datasets Prepration

• Market1501
• DukeMTMC-reID
• MSMT17

Train and Evaluation

# a demo on DukeMTMC-reID dataset
# please see more in ./configs
# train BagTricksIBN50 as teacher model
CUDA_VISIBLE_DEVICES=$CUDA python ./tools/train_net.py --config-file ./projects/DistillReID/configs-bagtricks-ibn-dukemtmcreid/bagtricks_R50-ibn.yml MODEL.DEVICE "cuda:0"
# train BagTricksIBN18 as student model 
CUDA_VISIBLE_DEVICES=$CUDA python ./projects/DistillReID/train_net.py --kd --config-file ./projects/DistillReID/configs-bagtricks-ibn-dukemtmcreid/KD-bot50ibn-bot18ibn.yml MODEL.DEVICE "cuda:0"

Experimental Results and Trained Models

DukeMTMC-reID

Rank-1 (mAP) / Q.Time/batch(128)		Student (BagTricks)
		IBN-101	IBN-50	IBN-34	IBN-18
Teacher (BagTricks)	IBN-101	90.8(80.8)/0.3395s	90.8(81.1)/0.1984s	89.63(78.9)/0.1760s	86.96(75.75)/0.0854s
	IBN-50	-	89.8(79.8)/0.2264s	88.82(78.9)/0.1761s	87.75(76.18)/0.0838s
	IBN-34	-	-	88.64(76.4)/0.1766s
	IBN-18	-	-	-	85.50(71.60)/0.9178s

Market-1501

Rank-1 (mAP) / Q.Time/batch(128)		Student (BagTricks)
		IBN-101	IBN-50	IBN-34	IBN-18
Teacher (BagTricks)	IBN-101	95.43(88.95)/0.2698s	95.19(89.52)/0.1791s	94.51(87.82)/0.0869s	93.85(85.77)/0.0612s
	IBN-50	-	95.25(88.16)/0.1823s	95.13(87.28)/0.0863s	93.44(82.92)/0.0621s
	IBN-34		-	94.63(84.91)/0.0860s
	IBN-18	-	-	-	92.87(81.22)/0.0615s
Average Q.Time		0.2698s	0.1807s	0.0864s	0.0616s

MSMT17

Comming Soon

Contact

This project is conducted by Guan'an Wang (guan.wang0706@gmail) and Liao Xingyu.