furkanom / tf-blazeface Goto Github PK

which is the format of this Input data set for training?Can you give the data preparation process?Give an example. Thanks!

First of all, thank you for sharing your code :)

I have been trying to get the the300w_lp but it is giving errors using the way you call it. I also investigated what the error could be as in https://stackoverflow.com/questions/63532993/how-to-load-downloaded-image-dataset-the300w-lp-in-tensorflow, it only says here is a list of available TF datasets. It is like there is no TF dataset as the300w_lp.

Could you please elaborate?

I also tried to convert other datasets to train the network but, unsuccessful.

Hi, how can I train the model with custom dataset ( the ground truth is in CSV Format) ?

Hello
How are you?
I've trained a model for detecting 6 keypoints and face bounding box.
I used your project without any modification.
I tested the model on some images but it's accuracy is very poor.

Even for some images, it is impossible to detect the face bounding box.
Please let me know the reason.
Thanks

should the line 69 and 70 be swapped?

gt_boxes = generate_bboxes_from_landmarks(gt_landmarks)
gt_landmarks = filter_landmarks(gt_landmarks)

The bounding box should be the bbox of filtered landmarks rather than the whole 68 landmarks, is it?

Hello
how are you?
Thanks for contributing this project.
I have a question.
Is it possible to detect more 6 points for landmark even if the Blazeface architecture is changed ?
Thanks.

I tried use weights from mediapipe .tflite, but they have different kernel size (3,3). Are there any trained model to download and check if it's working?

Hey, thanks for the implementation. Can we use provided an official BlazeFaze model to do test inference?

add the following line.
actual_deltas = tf.where(tf.expand_dims(pos_cond, -1), actual_deltas, tf.zeros_like(actual_deltas)).
for prior_boxes are not all zeros.
the clear zero action before bbox_utils.get_deltas_from_bboxes_and_landmarks was wasted.