tensorflow_person_image_remove_background

• Model weights url
• Object Detection: SSD ResNet50 V1 FPN 640x640 (RetinaNet50)
• Image Segmentation: Unet

Target

• Remove person images or videos background
• 此模組適合正面且單人的圖片
• This model is suitable for frontal and single-person images.

Using

• Python
• Tensorflow
• Opencv
• Numpy
• and so on.

What our code do?

1. Object Detection > Get boxes
2. Image Segmentation in each detected image boxes > Get mask
3. Remove background > Result

You can try this model.

• https://youtu.be/C9AffviLrxo
• https://colab.research.google.com/drive/1rDcVczczKy8IbUnfA4aAVrM_AKSwuJl7

References

• https://www.tensorflow.org/?hl=zh-tw
• https://github.com/tensorflow/models
• https://github.com/tensorflow/models/blob/master/research/object_detection/colab_tutorials/eager_few_shot_od_training_tf2_colab.ipynb