ZHEGG

Results 10 issues of ZHEGG

请问一下为什么evaluation的时候显存一直在一点点增加呢,最后就爆显存了。

Thank you for releasing the nice code! I try e2e_trie but not use GT_bboxes for inference and have this problem when computing the f1-score ![image](https://user-images.githubusercontent.com/65839993/188942834-35d3658a-a26a-49a5-816a-b591b6e46df6.png) I think the number of...

车票数据好像只包含合成的300k和测试的400张真实

您好!我在使用maskrcnn的config作用于wildreceipt数据集时,config文件如下 """ #################################################################################################### # Copyright Info : Copyright (c) Davar Lab @ Hikvision Research Institute. All rights reserved. # Filename : mask_rcnn_r50_r32_e2e_finetune_ic13.py # Abstract : Model settings for mask rcnn...

I try to install this environment and skip the line[26:] which is responding to tp&east postprocessing and wrap-ctc module but a new error occur and I found my MMCV CUDA...

在使用centernet的代码复现其physical structure中的P,R,F指标时,发现自己的结果为 ![1](https://user-images.githubusercontent.com/65839993/180694560-c198e1e3-1ce2-41b1-b8bf-d8c3e3effde4.JPG) 在iou=0.9时检测结果为0 并可视化了一些结果,发现其大概是正常的 ![2](https://user-images.githubusercontent.com/65839993/180694877-560e816e-ba1a-43e5-ad54-4f9693e7fe9a.png) 所以对您iou=0.9的指标有些疑惑 我使用了3090x4来复现您的结果其opt如下 ==> torch version: 1.7.0+cu110 ==> cudnn version: 8004 ==> Cmd: ['src/main.py', 'ctdet', '--exp_id', 'wtw_lr_1024', '--dataset', 'wtw', '--batch_size', '32', '--master_batch', '9', '--lr', '1.25e-4',...

![image](https://github.com/LlamaFamily/Llama-Chinese/assets/65839993/6fbc0dee-b078-4475-a555-773214bdd505) 无法理解这是为什么

I noticed that your paper used mask-rcnn but I found dataset dont have Polygon annotation, how did you use mask-rcnn? Thanks!

我想查看transformerdecoderlayer里的attn,在主函数中get_local.activate(),但是cache中的['transformer_decoderlayer']中是[],我的模型结构是train中调用model,model中调用transformer

![image](https://github.com/meta-llama/llama/assets/65839993/bb3b4a76-3d61-4b5d-8036-8e0b5df6e748)

model-access