how to speed up sequential dataset

[ericufatass]

Thanks for your amazing job. I have tried the demo, it did well. When I replace the datapath in demo file with my own dataset(which is built by slicing pics from a video, about 2000 imgs) ,the full pipeline cost a lot. Is there any way to speed up for my sequential dataset ? Or can u give me some advised param setting ? Thanks a lot🥹

[Phil26AT]

Hey @ericufatass

2000 images is quite a lot, and the demo matches pairs exhaustively, and you end up with over 1M pairs to match (thats completely unfeasible :D). Instead, you should create a pair shortlist, either based on image retrieval (only match each image to its 50 closest neigbors) or temporal information (try this PR, or one of the solutions listed there).