github
/
darknet
mirror of https://github.com/AlexeyAB/darknet.git
1
0
Fork 0
Code Issues Releases Wiki Activity
1724 Commits
1 Branch
7 Tags
380 MiB
Tag: Branch: Tree: 02e7109b8b
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from '02e7109b8b'
${ noResults }
Commit Graph

87 Commits (02e7109b8baae64f65b0fd35c49ef924863580e0)

Author SHA1 Message Date
AlexeyAB d11caf486d Fixed BiFPN and label_smothing for Detection
5 years ago
AlexeyAB 5e54523960 Added loss_scale for Mixed-Precision on Tensor-Cores
5 years ago
AlexeyAB c814d56ec1 Added [net] dynamic_minibatch=1 for increasing mini_batch_size when random=1 is used
5 years ago
AlexeyAB ef979a1fd2 For CPU and GPU structures in the darknet.h have the same size
5 years ago
AlexeyAB 6fb817f68b Accelerated DropBlock on GPU
5 years ago
AlexeyAB 64fb042c63 Averaging Mean/Variance for several mini_batches inside 1 batch. Path to CBN.
5 years ago
vinjn 3000c3111f Expose crop_image and resize_min as LIB_API
5 years ago
AlexeyAB 298805cefe Added new [convolutional] activation=normalize_channels_softmax_maxval for ASFF
5 years ago
AlexeyAB 4b59770886 Added max_delta=10 or 1 to the [yolo] and [Gaussian_yolo] layers, to limit delta and reduce the probability of Nan.
5 years ago
AlexeyAB f1643cbf2e Added: weights_normalizion = relu or softmax for [shortcut] layer
5 years ago
AlexeyAB a66bc8d6ff Resizing Detector: random=1 (/1.4 - x1.4), random=3 (/3.0 - x3.0), random=0 (no resizing). Added param: resize_step=32 to the [net] section: use resize_step=128 for 7 subsampling layers.
5 years ago
AlexeyAB 9bd88d7fd7 weighted [shortcut] layer
5 years ago
AlexeyAB b267e34487 multilayer [shortcut] in progress
5 years ago
AlexeyAB 489ea7fdce Minor fix
5 years ago
AlexeyAB 2c08a20b22 Added flag -benchmark_layers for: classifier demo, detector test, detector train
5 years ago
AlexeyAB 114a7f942b Added command line param -dontuse_opencv for training Classifier. Also use GaussianBlur instead of bilateralFilter for blur=1 for training Classifier and Detector.
5 years ago
AlexeyAB af54d95184 Added stretch_sway
5 years ago
AlexeyAB 005714a10c Added [local_avgpool] layer
5 years ago
AlexeyAB e33ecb785e Added param counters_per_class=100,200,... <for each class> to [yolo] and [Gaussian_yolo] to solve class-imbalance-problem
6 years ago
AlexeyAB 1626f9d495 Some fixes
6 years ago
AlexeyAB 5e035565ff Rotate conv-kernel
6 years ago
AlexeyAB a08c872564 Sway conv-kernel
6 years ago
AlexeyAB 642c065c0e Fixed DropBlock [dropout] dropblock_size_abs=7 dropblock_size_rel=0.5
6 years ago
AlexeyAB 1df3ddc7d6 Added DropBlock. Use [dropout] dropblock=1 dropblock_size=0.5 probability=0.1
6 years ago
AlexeyAB 2a873f3448 Added label_smooth_eps=0.1 for [net] layer for Label Smoothing for Classifier
6 years ago
AlexeyAB 318919e1cb Added label_smooth_eps=0.1 for [yolo] and [Gaussian_yolo] layer for Label Smoothing
6 years ago
AlexeyAB c9c745ccf1 Added [convolutional] activation=normalize_channels_softmax
6 years ago
AlexeyAB 68db8efb27 Added Fusion features for ASFF and for BiFPN(EfficientDet)
6 years ago
AlexeyAB 0d98f20be1 You can train with mini_batch=128 (batch=256 subdivisions=2) on GPU 8 GB VRAM + 128 GB CPU-RAM and more. Set in cfg-file: optimized_memory=3 and workspace_size_limit_MB=2000 or 4000
6 years ago
AlexeyAB 48914023da Added init_cpu() function to init CPU just to speedup the 1st detection (if GPU=0)
6 years ago
AlexeyAB 8cb3ee4e79 diounms_sort() fixed
6 years ago
AlexeyAB 4f70fc14e9 Added DIoU and CIoU to [yolo] and [Gaussian_yolo]
6 years ago
AlexeyAB 6e5bdf1282 Bug fix
6 years ago
AlexeyAB e7e85b358c Added delta_yolo_box_accumulate(). Added iou_thresh=0.213 parameter to [yolo] layer IoU(anchor,truth).
6 years ago
AlexeyAB ded620ac6b fixed memory deallocation
6 years ago
AlexeyAB d91d59a22f Optimized memory allocation for Detection (inference only), without allocation memory for training
6 years ago
AlexeyAB 0fa9c8f105 Added groups= and groupd_id= params to the [route] layer
6 years ago
AlexeyAB bf8ea4183d Added MISH activation, use activation=mish in [convolutional] layers
6 years ago
AlexeyAB b3a2495298 Added Gaussian YOLOv3 layer [Gaussian_yolo]
6 years ago
AlexeyAB be5d0d6693 Added assisted_excitation=1 for [convolutional] layer on GPU
6 years ago
AlexeyAB 11b8b2bddf Added antialiasing=1 param for [convolutional]-layer on GPU
6 years ago
AlexeyAB 8d80a65288 Max pool layer can use stride=2 or stride_x=2 stride_y=4 (isn't tested well)
6 years ago
AlexeyAB c1f184249e Added assisted_excitation_forward for training on CPU
6 years ago
AlexeyAB 2fa9792c56 Added [sam] layer
6 years ago
AlexeyAB 54e2d0b0e8 Added ability to use letter_box resizing in darknet.py sample
6 years ago
AlexeyAB d4402d29c2 Added maxpool_depth= and out_channels= params to [maxpool]
6 years ago
AlexeyAB 88cccfcad4 Added Swish-activation
6 years ago
AlexeyAB c9129c2078 Added letter_box=1 param in [net] section (cfg-file) for keeping aspect ratio during training
6 years ago
AlexeyAB cc41339805 Added [scale_channels] layer for squeeze-and-excitation blocks
6 years ago
AlexeyAB bd689f91ac Added TridentNet (state-of-art on MS COCO)
6 years ago