TypeError: loop of ufunc does not support argument 0 of type Entity which has no callable conjugate method

[zhenzi0322]

import towhee


class Resnet50:
    def test(self):
        print(dir(towhee))
        feat = towhee.glob['path']('../tmp/data/01/*.jpg').head(2).image_decode['path', 'img']() \
            .image_embedding.timm['img', 'vec'](model_name='resnet50') \
            .tensor_normalize().to_list()
        print(feat[0], type(feat[0]), feat[0].path, type(feat[0].path))
        print(dir(feat[0]))


if __name__ == '__main__':
    b = Resnet50()
    b.test()

[shiyu22]

Hi @zhenzi0322, I think that you need to add schema for the tensor_normalize() operator, such as:

feat = towhee.glob['path']('../tmp/data/01/*.jpg').head(2).image_decode['path', 'img']() \
            .image_embedding.timm['img', 'vec'](model_name='resnet50') \
            .tensor_normalize['vec', 'vec']().to_list()

[shiyu22]

You can use select:

feat = towhee.glob['path']('../tmp/data/01/*.jpg').head(2).image_decode['path', 'img']() \
            .image_embedding.timm['img', 'vec'](model_name='resnet50') \
            .tensor_normalize['vec', 'vec']().select['path', 'vec']()

[zhenzi0322]

How can I retrieve all the paths and all the vec in this way?

[zhenzi0322]

TypeError: Object of type ndarray is not JSON serializable

feat = towhee.glob['path']('../tmp/data/01/*.jpg').head(2).image_decode['path', 'img']() \
            .image_embedding.timm['img', 'vec'](model_name='resnet50') \
            .tensor_normalize['vec', 'vec']().select['path', 'vec']()
        feat_list = feat.to_list()  # [<Entity dict_keys(['path', 'vec'])>, <Entity dict_keys(['path', 'vec'])>]
        print(feat_list[0])  # TypeError: Object of type ndarray is not JSON serializable

[shiyu22]

It works for me, I think it's an error caused by fastAPI because you need to return a string value.

[zhenzi0322]

How to convert it to 512dim, the default converted dim is 2048