Tracking of mice in novel object recognition The tracking was done using DeepLabCut
The performance pf the network was suboptimal when I use only one video for labelling and training the network
I cut 30 sec from the middle of each video and concatenated them together in one video and extracted the frames and trained the network on this video and the performance improved dramatically