Owen

output face with anime outline

1

some outputs that your model `Face Portrait v2` generated have a very good anime outline :   and   How to do it ? I tried to redo...

I found the colorization model likes color the object by red color!

3

when I color some photos,I found the colorization model likes color the object by red color! for example:    There's something wrong in it !

Training Network for Pedestrian Detection

1

does anyone train the PyTorch-YOLOv3 on pedestrian detection data sets (e.g., Caltech)? or train the network just for VOC or COCO datasets 's person class ?

UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead.

6

when I train yolo v3 on coco, it WARNING that "/pytorch/aten/src/ATen/native/IndexingUtils.h:20: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead." It's so many! How to...

Edit face in Eyeglasses, Gender, Hair Color, Pose , Smile ... direction failed

2

I edited face in Eyeglasses, Gender, Hair Color, Pose , Smile ... direction by MODEL_ZOO/StyleGAN2/[ffhq-1024x1024](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155082926_link_cuhk_edu_hk/EX0DNWiBvl5FuOQTF4oMPBYBNSalcxTK0AbLwBn9Y3vfgg?e=Q0sZit&download=1)   it shown that some attributes/directions are highly correlated with each other .

something wrong with the groundingdino detect result

6

Hi, thank you for your excellent work. So quick ! I tested, and I found something wrong with the groundingdino detect result . I show it : Input image: ![13201678780389_...

运行的结果没有示例图中好

7

非常棒的工作，标注效果相比Blip的有了很大的提升！nice！  主业的这张图中RAM的结果中如你展示和提醒的是有`lamp和door`标签的，但是我跑出来的结果中却没有  是什么原因导致的呢？

我在本地测了一下去固定位置的水印，但是效果感觉比你的差一些，有什么trick？

2

很惊艳的工作！所以本地测试了一下，发现同类型的视频感觉比你的demo差一些 源视频[hild-runs-barefoot-on-grass-park-joyful](https://www.shutterstock.com/zh-Hant/video/clip-1100574563-child-runs-barefoot-on-grass-park-joyful)跟你的demo一样来自**shutterstock**，下载后的视频默认是webm格式的，我用FFmpeg转为了基于libx264的MP4格式 https://github.com/sczhou/ProPainter/assets/21029719/35e44553-9789-40b9-a1b2-24af7ae8ae51 mask是直接使用的你的mask，因为视频尺寸都是一致的 但是结果相比你demo中的，感觉要差很多，mask区域有比较明显的模糊块 结果视频为： https://github.com/sczhou/ProPainter/assets/21029719/81fb9a4b-d8f5-43d2-8557-e203b03b68ac 有什么提升的办法？

input tensor count changed from 5 to 15, will compile again.

6

### Describe the bug When trying to accelerate ControlNet inference based on Stable Diffusion XL using OneDiff in ComfyUI, I found that modifying the end_percent parameter of the ControlNet module...

### Describe the bug When I was testing the acceleration inference of DeepCache, I encountered an error when I modified the inference size to 720*960. However, when I modified the...