验证集的iou值太低了

[sijie-Xu]

I have modified the code for my own remote sensing data and used bounding box prompts for guidance, but during training, the IoU on the validation set is extremely poor.
我修改成自己的遥感数据,并使用框提示进行提示.但我在训练的过程中验证集的iou非常的差

I also modified the visualization part, displaying results using bounding box prompts, and converted the predicted values to 255 for visualization purposes
我还修改了可视化的部分，显示了通过框提示的结果，并将预测值转化成255以便可视化

But there is at least some overlap in the visualization, indicating that the IoU is not so low
但是在可视化中至少有重叠的部分，至少说明iou不至于这么低

Could it be due to the handling of multiple thresholds?
请问是否是因为多个阈值处理的问题

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Title: The iou value of the validation set is too low

[WuJunde]

can you pls share your IOU calculation code?

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

can you please share your IOU calculation code?

[sijie-Xu]

can you pls share your IOU calculation code?

I didn't change the code for the calculation of the IoU. Whether it's because the number of training epochs is too small, I've only trained for 14 epochs.

[WuJunde]

can you pls share your IOU calculation code?

I didn't change the code for the calculation of the IoU. Whether it's because the number of training epochs is too small, I've only trained for 14 epochs.

Try training over 100 epochs

[JaireYu]

@sijie-Xu #135 may help, please check~