宠物检测

识别指定区域的猫和狗 注： 1.画面中的猫狗不能和背景颜色太过接近（比如黑天角落有个黑狗，此种场景人眼也难以识别） 2.猫和狗均没有品种限制 3.猫狗团成球的情况也不支持 精度对应场景或者限制条件 1）场景限定：家庭、院落、街道、厨房和餐厅等场景 2）图片分辨率：1080p 3）设备/镜头：支持RGB可见光 4) 识别对象尺寸：最小尺寸为150*150px；遮挡<30% 5）识别的种类：猫和狗Identify cats and dogs within specified regions. Notes: 1. Cats and dogs in the input image must not be overly similar in color to the background (e.g., a black dog in a dark corner, a scenario that is also difficult for human beings to recognize). 2. There are no breed restrictions for either cats or dogs. 3. Scenarios where cats or dogs curl into a ball are not supported. Accuracy corresponds to the following scenarios and constraints: 1. Scene limitations: scenarios including homes, yards, streets, kitchens, dining rooms, etc. 2. Image resolution: 1080p 3. Device/lens: supports RGB visible light 4. Size of recognized objects: minimum size is 150×150px; occlusion rate < 30% 5. Recognized categories: cats and dogs

搜集汇总

背景与挑战

背景概述

该数据集是一个算法模型，用于识别指定区域的猫和狗，适用于家庭、院落、街道、厨房和餐厅等场景，要求图片分辨率为1080p，识别对象最小尺寸为150*150px，遮挡小于30%。