yifanxu74 / mq-det Goto Github PK

作者您好，我对您的工作非常感兴趣，可惜我没有object365数据集，无法顺利debug代码，请问我能否用其他数据集代替呢？

Does the current release code support groundingdino training? And what is the training script like?

Hi, great work!
Do you also have zero-shot evaluation results on COCO?

Hello! In the path catalog I see that you use a clean version of Objects365 for grounding pre-training here. I wonder how do you built it. There is no mention on the DATA.md as far as I'm concerned. Thanks for your answer!

请问这个预训练模型是从哪里获取，可以分享下吗QUERY_BANK_PATH: 'MODEL/object365_query_5000_sel_tiny.pth'

While torch.load('mq-glip-l') , there is a RuntimeError: "PytorchStreamReader failed reading zip archive: not a ZIP archive". mq-glip-t is the same as well. mq-glip-l and mq-glip-t are downloaded from https://drive.google.com/file/d/1O_eb1LrlNqpEsoxD23PAIxW8WB6sGoBO/view and https://drive.google.com/file/d/1n0_D-tisqN5v-IESUEIGzMuO-9wolXiu/view

Hi Author,

Great work! I'm wondering if you have any plan on developing a demo showing few-shot ability of the model.

Specifically, we can prepare some vision queries (a few examples with novel classes). And with these vision queries, we perform object detection of an input image from the same domain as the vision queries. I'm excited to seeing the performance of MQ-Det in this scenario. (Since, I blieve that with VisionAndText queries, MQ-Det can definitely perform better! :))

The function I propsed is very similar to the playground demo of OWL-ViT (link below).

https://colab.research.google.com/github/google-research/scenic/blob/main/scenic/projects/owl_vit/notebooks/OWL_ViT_inference_playground.ipynb

您的论文中GLIP-T在OdinW13上zero-shot的AP值和GLIP论文中不一致，是有设置上的差异吗
MQ-DET中

GLIP论文中

作者您好，在按照readme执行训练mq-groundingdino-t.yaml时，出现keyerror: positive map。求助，谢谢！

Hi, great work!
Do you have an inference code for a single image?

物体检测比较清楚，但我一直不太明白结合 “in the wild” 是什么意思，您能否根据文章解释一下？谢谢。

While finetuning following the instruction , an error "too many values to unpack" happens in "for iteration, (images, targets, idxs, positive_map, positive_map_eval, greenlight_map) in enumerate(data_loader, start_iter):" (line 92 of maskrcnn_benchmark/engine/trainer.py). What is this problem?
The command is "python tools/eval_odinw.py --config_file configs/pretrain/mq-glip-l.yaml --opts 'MODEL.WEIGHT MODEL/mq-glip-l' --setting 5-shot --add_name large --log_path 'OUTPUT/custom_log/' --task_config configs/custom_dataset.yaml"