mirror of
https://github.com/deepseek-ai/DeepSeek-VL2.git
synced 2026-11-05 01:07:23 +00:00
[GH-ISSUE #94] How to correctly visualize bounding box? 请问如何正确可视化边界框？ #77
Labels
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
deepseek-ai/DeepSeek-VL2#77
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @CS-FantaC on GitHub (Feb 21, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-VL2/issues/94
Simple Inference Example 里的代码:
conversation = [
{
"role": "<|User|>",
"content": "\n<|ref|>The giraffe at the back.<|/ref|>.",
"images": ["./images/visual_grounding_1.jpeg"],
},
{"role": "<|Assistant|>", "content": ""},
]
模型输出为[580, 270, 999, 904],直接把bbox画在原图上发现有偏移，如图所示，请问有可视化bbox的代码吗？]我认为应该是模型给出的坐标不是原图分辨率对应的坐标，而是变换后的图像中的坐标。请问具体经过了什么裁剪变换呢？
@CS-FantaC commented on GitHub (Feb 22, 2025):
后面发现只是因为Tiny model定位不准，small model输出的boundingbox就可以准确定位。
@hengrui0516 commented on GitHub (Feb 24, 2025):
想请问下[580, 270, 999, [904]这种输出的嵌套列表是什么意思呢，还有的时候模型会输出奇数个值该怎么处理呢
@CS-FantaC commented on GitHub (Feb 24, 2025):
抱歉，我打字时打多了一个[，并不是嵌套列表。奇数个数值肯定是错误输出，只能让模型重新输出