coco

How to train a street scene and put it into the 3d software?

4

I just read the [nerf_dataset_tips](https://github.com/NVlabs/instant-ngp/blob/master/docs/nerf_dataset_tips.md) and find it in the last section: "The colmap2nerf.py script assumes that the training images are all pointing approximately at a shared point of interest,...

What if I only want to train the human nerf part, what should I do, and how long it would take for training the human part?

1

I tried the [humannerf](https://github.com/chungyiweng/humannerf), but I don't know why something went wrong [my issue](https://github.com/chungyiweng/humannerf/issues/49), so I'm trying to use humannerf in ml-neuman to train the only human part, can you...

How to get Da-pose canonical human from a single image? And use the canonical human to do what?

11

Use the key points and bounder of human in the image, and compare it with SMPL mesh to train the parameters in SMPL for a suitable “meta human”？Then control the...

paddleocr v3的rec的"训练模型"是仅做了初始化吗？

1

paddleocr v3的rec的"训练模型"是仅做了初始化吗(只保存了模型结构？因为我在train的时候发现acc为0，在大概7个epoch达到0.5+，如果是已经在某些数据集上pretrained了话acc应该不会是0吧，但是“best_accuracy”的文件名又有点儿奇怪像是做了pretrained了一样

"metadata.json" and "--type movement"

1

Thanks for your marvellous research!!! I still have some questions: What are these two things used for? `metadata.json` in ``` monocular ├── images │ └── ${item_id}.png ├── masks │ └──...

Human legs disappear when rendering?

8

I try to render novel poses according to #19 , but when I set ```python dst_poses[16*3+0] = 1.0 # +57deg dst_poses[17*3+0] = -1.0 # -57deg ``` the human legs disappeared,...

如果想绕过deepspeed做finetune，可以在train的时候直接model.step()来实现吗？

2

 或者有什么办法（或者需要注意修改哪些地方），才能实现解开对deepspeed的依赖呢？

怎么做到把gltf压缩到6kb而且质量这么清晰呢？图片转gltf太大了，至少几百kb

According to Hungarian algorithm, we can get the bipartite graph maximum matching between ground truth bounding boxes and proposals, that means we can find a proposal for each gt? If...

so the implement seems different between diffusiondet and detr, any significant change about the macher module?