郑启航

which board did you use ? my code and kendryte demo code only support kendryte KD233 board. The DVP driver will auto convert rgb565 image to rbg888 format and put...

yes. The facedetect_landmark is realtime demo. The yolo demo just infer once, you can custom image array.

Just like the classical face recognition task, make the batch data input, the `y_true` is reference image, compute the each data cos similarity with `y_true` , if current data have...

In order to speed up the loss calculation, I set `batch_idxs` to a constant value. ```python if batch_size: self.batch_size = batch_size self.batch_idxs = tf.expand_dims( tf.range(0, batch_size, dtype=tf.int32), 1) # shape...

1. for image classification: I use code as following ensure y_pred as `cos similarity`. ```python kl.Lambda(lambda x: tf.nn.l2_normalize(x, 1), name='emmbeding'), kl.Dense(10, use_bias=False, kernel_constraint=k.constraints.unit_norm()) ``` 2. yse, But if k >...

I think that the optimization goal set by `circle loss` cannot be achieved, so his loss gradient will be quite large. At the same time, the `ap` and `an` is...

you can. but the pretrained backbone for k210 I forced the deletion of some layers, so maybe you need retrain backbone.

tiny-yolo v3 is too big , so I use mobilenetv1-yolov3 to fit k210.

My code test in tensorflow r1.14, maybe you should update your tensorflow. All package: numpy==1.16.2 tensorflow_gpu==1.14.0 opencv_python==4.0.0.21 matplotlib==3.0.3 Pillow==6.1.0

It suddenly failed again, but I find a way to avoid this bug, we can custom a code-workspace file. 1. encode your remote container config setting: ```sh echo -n '{"containerName":"/your_docker_name","settings":{"host":"ssh://[email protected]"}}'...