search

deepalchemist / deep-person-search

deep learning person search in PyTorch.
73 stars 22 forks source link

Evaluation: 0.0% result #1

Closed Roberto7766 closed 4 years ago

Roberto7766 commented 4 years ago

Hi, during evaluation with default values I obtain the following result:

Experiment dir: /home/roberto/PersonReID/deep-person-search/20201004/test Loaded dataset ssm_test for testing 6978 roidb entries Backbone loading ImageNet pretrained weights from ./cache/pretrained_model/resnet50_caffe.pth Engine [baseline] was created Not loading checkpoint!

ModelParams : 35.05 M

extracting probe features ... im_exfeat: 2900/2900 0.008s
extracting gallery features ... total time:1.297s, reid time:0.002s
evaluating detections all detection: recall = 0.00% ap = 0.00% /home/roberto/anaconda3/envs/dps/lib/python3.6/site-packages/sklearn/metrics/ranking.py:528: RuntimeWarning: invalid value encountered in true_divide recall = tps / tps[-1] labeled_only detection: recall = 0.00% ap = nan% search ranking: recall = 0.00% mAP = 0.00% top- 1 = 0.00% top- 5 = 0.00% top-10 = 0.00% test time: 36.9min

The command used is:

_CUDA_VISIBLE_DEVICES=0 python main.py --is_test --benchmark ssm --eval_batch_size 5 --backbone bsl --in_level C5 --clstype oim

It is quite a strange result. What can be wrong? Thanks

deepalchemist commented 4 years ago

Firstly, download the trained model to path_to_the_checkpoint. Secondly, load the checkpoint by adding --load_ckpt "absolute_path_to_your_checkpoint" to your command.