How to train the CNN? - Githubissues

Grabber commented 5 years ago

CNN weights are set as static variables. Is there any code available for training in new datasets?

ShiqiYu commented 5 years ago

I just upload the caffe model for training.

Grabber commented 5 years ago

@ShiqiYu

Is there any Caffe modification to train using the network spec files? How do you train it, step by step?

  multibox_loss_param {
    loc_loss_type: L2
    conf_loss_type: SOFTMAX
    loc_weight: 1.0
    num_classes: 2
    share_location: true
    match_type: PER_PREDICTION
    overlap_threshold: 0.34999999404
    use_prior_for_matching: true
    background_label_id: 0
    use_difficult_gt: true
    neg_pos_ratio: 3.0
    neg_overlap: 0.34999999404
    code_type: CENTER_SIZE
    ignore_cross_boundary_bbox: false
    mining_type: MAX_NEGATIVE
  }

ShiqiYu commented 5 years ago

Use Nvidia Caffe

Luiz Vitor Martinez Cardoso notifications@github.com于2019年3月16日周六05:29写道：

@ShiqiYu https://github.com/ShiqiYu

Is there any Caffe modification to train using the network spec files? How do you train it, step by step?

multibox_loss_param { loc_loss_type: L2 conf_loss_type: SOFTMAX loc_weight: 1.0 num_classes: 2 share_location: true match_type: PER_PREDICTION overlap_threshold: 0.34999999404 use_prior_for_matching: true background_label_id: 0 use_difficult_gt: true neg_pos_ratio: 3.0 neg_overlap: 0.34999999404 code_type: CENTER_SIZE ignore_cross_boundary_bbox: false mining_type: MAX_NEGATIVE }

— You are receiving this because you were mentioned.

Reply to this email directly, view it on GitHub https://github.com/ShiqiYu/libfacedetection/issues/99#issuecomment-473447226, or mute the thread https://github.com/notifications/unsubscribe-auth/AG0eHQw_6mVDWl4T_5mu5HCcb7j-hdQEks5vXBDGgaJpZM4b2WPj .

ShiqiYu / libfacedetection

How to train the CNN? #99