You Zhiyuan
You Zhiyuan
I'd like to verify that: 1. whether your images have GT (density map) ? 2. how your exemplars are annotated ? Each image has exemplars annotated by boxes, OR, all...
1. If no GT (density map), you should remove the codes about "density" in `./datasets/custom_dataset.py`. 2. It is better for each image to have exemplars annotated by boxes. In this...
> I see. I assume then the output will be my input image but with the detected object segmented and the count? Yes, you are right.
Thanks for your reply. Also, thanks again for your excellent work, which provides a new aspect to understand object detection.
Hi~ I have checked your email and both attached config files. **_Currently, I could not get the original config files and checkpoints, since I have left SenseTime._** I want to...
Actually, I did not do any special design for different resnets. The resnet18 & resnet34 use **_BasicBlock_**, while the resnet50 & resnet101 use **_Bottleneck_**. Since the resnet is loaded with...
Another difference is the number of GPUs. I used 8 GPUs for training, which means `batch_size=64`. How about increasing the batch_size ?
Because the performance gap is relatively large, the only element that could cause such a difference is the feature layer. So you could have a try with `outlayers: [1,2,3,4], outlayers:...
Hi~ I do not have ideas about the distortions of the photos taken by Apple 12. You could have a try.
No. I finally use the in-painting model provided by Stable Diffusion.