Closed ChrisAllenMing closed 6 years ago
Sorry, we change the batch size from 1 to 4 so that the reshape function get confused on operating on the cell matrix.
Hello, I met this problem, too. This code works only if the batch size is 1. Have you got the solution to train the network in a bigger batch size?
When we train your model WSDDN, we met a problem on reshaping the matrix of confidence on proposals (i.e. imdb.images.boxScores{batch}). The error logs are listed as belows.
`Error using reshape Size arguments must be integer scalars.
Error in cnn_wsddn_train>getBatch (line 204) boxScore = reshape(imdb.images.boxScores{batch},[1 1 1 numel(imdb.images.boxScores{batch})]);`