Question about Caltech dataset

[zsyawn]

Hello Doctor, I have a question to ask you, when evaluating the Caltech dataset, you are using the Caltech new annotaions provided by Shanshan Zhang, The format is .txt, and the caltech matlab evalutation tool uses the original annotations and the format is .vbb, so how can I use this new annotation for Evaluation? I need your help, thank you very much!

[hasanirtiza]

Yes, we use the new annotations for training. We have added the details on how to evaluate on Caltech along with the Matlab evaluation code here. In the evaluation code, for the reasonable setting, you do not require annotations for the test set, as they are already contained in the package. Follow the instructions in Datasets-PreProcessing.md.

[zsyawn]

@hasanirtiza For caltech datasets，how to evalute the results on ALL set and Heavy Occlusion set? We only have one ground truth file named gt-Reasonable.mat in Eval_caltech.

[hasanirtiza]

You can generate all gt-files from the annotations of Caltech. Nonetheless, we have added now gt-Reasonable, gt-Occ=heavy and gt-All in the repo. We have also updated gt-Reasonable (new annotations by Shanshan). By default, dbEval.m will evaluate on reasonable subset. In order to evaluate any other subset, on line 52 of dbEval.m you need to add this line to evaluate say on occ=heavy :
exps = exps(9);

[zsyawn]

Thank you very much for your answer, it is very helpful to me