问题描述
有人可以快速解释或确认我对什么的猜测:
can someone quickly explain or confirm my guess on what i.e.:
max_detections_per_class: 100
max_total_detections: 100
就我而言 ssdlite_mobilenet_v2_coco.config
在第 134 和 135 行.
at line 134 and 135.
从预测的原始输出来看，我的猜测是这些预测总是尝试"检测图像中的 100 个对象，而不管图像中的实际对象数量.假设只有一只猫，在我返回的原始预测数据中仍然会检测到 100 个对象.如果模型训练正确，当然应该只有一个得分高的预测.
From the raw output of predictions, my guess is that the predictions always "tries" to detect 100 objects in the image, despite the actual number of objects in the image. Let's say there is only one cat, there still will be 100 objects detected in my returning raw prediction data. If the model is trained right, of course there should be only one prediction with a high score.
这样对吗?谢谢！
推荐答案
是的，你是对的.它将尝试检测 100 个对象.
Yes you are correct.It will try to detect 100 objects.
然后会对这 100 个检测进行分类，其中只有一个应该被正确识别为猫.但这也取决于您的非最大抑制配置.如果 NMS 具有低分数阈值和高 IoU，它可以显示猫的多次检测(我的意思是重叠检测).
Those 100 detections will be then classified, where only one should be correctly identified as a cat.But it also depends on your Non Max Supression config. If NMS has a low score threshold and a high IoU, it can show multiple detections for the cat (overlapped detections I mean).
你可以弄乱这些值，但从已发表的论文来看，每张图像的最大检测数应该总是比图像中的实际对象多约 3 倍.
You can mess with those values, but from the published papers, the number max detections per image should always be more ~3x more than the actual objects in the image.
根据我的经验，即使在我的数据中图像中的对象较少(例如每张图像 1 个)，我也能获得更好的结果.
In my experience I obtain better results leaving it as is, even if in my data I has less objects in the image (eg 1 per image).
这篇关于tensorflowjs 对象检测 coco ssd 配置文件 max_total_detections 解释的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！