How to generate zeroshot_triplet.pytorch?

Open waxnkw opened this issue 5 years ago • 0 comments

❓ Questions and Help

I try to use load_graph function in Scene-Graph-Benchmark.pytorch/maskrcnn_benchmark/data/datasets/visual_genome.py, and use the args like: train: load_graphs(roidb_file, split='train', num_im=-1, num_val_im=5000, filter_empty_rels=True, filter_non_overlap=False) val: load_graphs(roidb_file, split='val', num_im=-1, num_val_im=5000, filter_empty_rels=True, filter_non_overlap=False) test: load_graphs(roidb_file, split='test', num_im=-1, num_val_im=5000, filter_empty_rels=True, filter_non_overlap=False) and use code like union(val, test).difference(train) to calculate zeroshot triplets. However, I get 6,129 triplets for zeroshot triplets, while the zeroshot_triplet.pytorch you provide is 5,971. Is it available for your code of generating zeroshot_triplet.pytorch?

Dec 24 '20 09:12 waxnkw