I am running the command python tools/eval_net.py ucf101 1 flow FRAME_PATH \
models/ucf101/tsn_bn_inception_flow_deploy.prototxt models/ucf101_split_1_tsn_flow_reference_bn_inception.caffemodel \
--num_worker 4 --save_scores SCORE_FILE and getting an accuracy of 0.
I ran a similar command for the rgb stream and was getting good results with an accuracy of 93.18%.
I am doing everything on a cpu and have extracted flow images for only the first category i.e. ApplyEyeMakeUp and tested using the first test split weights.
I am running the command python tools/eval_net.py ucf101 1 flow FRAME_PATH \ models/ucf101/tsn_bn_inception_flow_deploy.prototxt models/ucf101_split_1_tsn_flow_reference_bn_inception.caffemodel \ --num_worker 4 --save_scores SCORE_FILE and getting an accuracy of 0. I ran a similar command for the rgb stream and was getting good results with an accuracy of 93.18%.
I am doing everything on a cpu and have extracted flow images for only the first category i.e. ApplyEyeMakeUp and tested using the first test split weights.