Zhuofan Xia

Hello @chenxinhe1 @huhulike, I will release a notebook in next a few weeks in [detection codebase](https://github.com/LeapLabTHU/DAT-Detection) to finish this.

The visualizations require a forward run of the object detection model on the MS-COCO validation dataset. Maybe I will release the visualization codes after the release of the object detection...

Hello @sharat29ag @wewan @cbq233333 @zzwei1, I will release a notebook in next a few weeks in [detection codebase](https://github.com/LeapLabTHU/DAT-Detection) to finish this.

Hello @hagianga21, Ideally, it is feasible to visualize the offsets in classification tasks. However, the images are very small so the visual effects could not be as promising as expected.

I have been working on polishing the codes, and am about to commit them in a few weeks.

Hello @tdchua @Filipwansen @Sairam13001 @stayhungry1 @yuanbianGit, Sorry for being so late. The code repos of detection and segmentation are updated along with the newer extended version of DAT. Please check...

> @X-Lai BTW, in utils/dataset.py line 269, why it is `return *data[0], inference` but not `return *data[idx], inference` ? @dengandong I find that the index passed into __getitem__ is not...

Hello @hhaAndroid, It is a common practice that one builds DINO or some recent strong detectors with the LARGE variant (~200MParams and ~40GFLOPs@224) pretrained on ImageNet-22K or larger datasets. However,...