PaddlePaddle
人体属性识别扩展其他的属性
首先，感谢大佬分享了PAR模型。我尝试了一下网站，效果相当不错。 https://ai.baidu.com/tech/body/attr
测试中我发现网站的输出类似
"person_info": [ { "attributes": { "orientation": { "score": 0.99993073940277, "name": "右侧面" }, "gender": { "score": 0.99729353189468, "name": "女性" }, "umbrella": { "score": 0.99989748001099, "name": "未打伞" }, "lower_color": { "score": 0.98367667198181, "name": "蓝" }, "face_mask": { "score": 0.99999898672104, "name": "戴口罩" }, "smoke": { "score": 0.99913018941879, "name": "未吸烟" }, "bag": { "score": 0.96256989240646, "name": "无背包" }, "upper_wear": { "score": 0.94788938760757, "name": "长袖" }, "is_human": { "score": 0.99996364116669, "name": "正常人体" }, "vehicle": { "score": 0.99995183944702, "name": "无交通工具" }, "glasses": { "score": 0.87400287389755, "name": "无眼镜" }, "headwear": { "score": 0.75554066896439, "name": "普通帽" }, "upper_wear_fg": { "score": 0.99537992477417, "name": "风衣" }, "upper_wear_texture": { "score": 0.9064017534256, "name": "纯色" }, "upper_cut": { "score": 0.99869328737259, "name": "无上方截断" }, "occlusion": { "score": 0.99969339370728, "name": "无遮挡" }, "lower_cut": { "score": 0.99461853504181, "name": "无下方截断" }, "cellphone": { "score": 0.99999898672104, "name": "看手机" }, "carrying_item": { "score": 0.98618221282959, "name": "无手提物" }, "age": { "score": 0.98111075162888, "name": "青年" }, "lower_wear": { "score": 0.99056619405746, "name": "长裤" }, "upper_color": { "score": 0.98910427093506, "name": "棕" } }]
比如打伞、口罩、衣服颜色、是否看手机、上方截断等相当丰富的信息，在PULC 人体属性识别模型里是没有的。 因此，想问一下多的这些属性是否提供了其他模型呢？ 谢谢！
暂时没有其他的模型。不过PULC训练非常简单，如果你有合适的数据集可以动手试试看。 https://github.com/PaddlePaddle/PaddleClas/blob/release/2.5/docs/zh_CN/training/PULC.md
暂时没有其他的模型。不过PULC训练非常简单，如果你有合适的数据集可以动手试试看。 https://github.com/PaddlePaddle/PaddleClas/blob/release/2.5/docs/zh_CN/training/PULC.md
好的谢谢。