search

KaihuaTang / Scene-Graph-Benchmark.pytorch

A new codebase for popular Scene Graph Generation methods (2020). Visualization & Scene Graph Extraction on custom images/datasets are provided. It's also a PyTorch implementation of paper “Unbiased Scene Graph Generation from Biased Training CVPR 2020”
MIT License
1.06k stars 229 forks source link

SGDet on Custom Images #190

Closed yayalvhai closed 1 year ago

yayalvhai commented 1 year ago

❓ Questions and Help

If I just detect scene graphs on one image, do I need to download the VG dataset?

yayalvhai commented 1 year ago

2023-02-09 21:19:54,192 maskrcnn_benchmark.data.build INFO: get dataset statistics... 2023-02-09 21:19:54,193 maskrcnn_benchmark.data.build INFO: Unable to load data statistics from: /home/user/projects/Scene-Graph-Benchmark/checkpoints/causal-motifs-sgdet/VG_stanford_filtered_with_attribute_train_statistics.cache

so the solution is to download the VG dataset ( 9GB+5GB) ?

ubuntufan4life commented 1 year ago

@yayalvhai Could you tell me how you solved the problem? Did you really have to download the VG dataset in the end?

yayalvhai commented 1 year ago

@yayalvhai Could you tell me how you solved the problem? Did you really have to download the VG dataset in the end

yeah. If just SGDet, you do not need to download VG datasets

zl535320706 commented 1 year ago

2023-02-09 21:19:54,192 maskrcnn_benchmark.data.build INFO: get dataset statistics... 2023-02-09 21:19:54,193 maskrcnn_benchmark.data.build INFO: Unable to load data statistics from: /home/user/projects/Scene-Graph-Benchmark/checkpoints/causal-motifs-sgdet/VG_stanford_filtered_with_attribute_train_statistics.cache

so the solution is to download the VG dataset ( 9GB+5GB) ?

After download Visual Genome image (108249 items), put it into datasets/vg/VG_100K and run bash scripts/test.sh, the program automatically generate VG_stanford_filtered_with_attribute_train_statistics.cache in checkpoints/PE-NET_PredCls/

@ubuntufan4life

majianbo3 commented 8 months ago

2023-02-09 21：19：54，192 maskrcnn_benchmark.data.build INFO：获取数据集统计信息. 2023-02-09 21：19：54，193 maskrcnn_benchmark.data.build信息：无法从以下位置加载数据统计信息：/home/user/projects/Scene-Graph-Benchmark/checkpoints/case-motifs-sgdet/VG_斯坦福大学_filtered_with_attribute_train_statistics.cache 所以解决方案是下载VG数据集（9 GB +5GB）？

下载Visual Genome图像（108249个项目）后，将其放入datasets/vg/VG_100K并运行bash脚本/test.sh，程序自动生成VG_斯坦福大学_filtered_with_attribute_train_statistics.cache in checkpoints/PE-NET_PredCls/

@ubuntufan4life

sorry,i dont know where /test.sh