# DnnLab
Dnnlab is a small framework for deep learning models based on TensorFlow.
It provides custom training loops for:
* Generative Models (GAN, cGan, cycleGAN)
* Image Detection (custom YOLO)
Additonaly custom Keras Layer:
* Non-Local-Blocks (Self-Attention)
* Squeeze and Excitation Blocks (SEBlocks)
* YOLO-Decoding Layer
Input pipeline functionality:
* YOLO (Tfrecords to Datasets)
* YOLO data augmentation
* Generative Models (Tfrecords to Datasets)
TensorBoard output:
* YOLO coco metrics (Precision (mAP) & Recall)
* YOLO loss (loss_class, loss_conf, loss_xywh, total_loss)
* YOLO bounding boxes
* Generative Models (Loss & Images)
## Requirements
TensorFlow 2.3.0
## Installation
Run the following to install:
```python
pip install dnnlab
```
Python库 | dnnlab-1.2.6.tar.gz
版权申诉
49 浏览量
2022-04-08
09:05:27
上传
评论
收藏 73KB GZ 举报
dnnlab-1.2.6.tar.gz （55个子文件） 
dnnlab-1.2.6 
MANIFEST.in 167B PKG-INFO 2KB LICENCE 11KB src dnnlab.egg-info PKG-INFO 2KB
requires.txt 120B SOURCES.txt 2KB top_level.txt 7B dependency_links.txt 1B dnnlab input_pipeline img_label.py 6KB yolo.py 9KB augmentation yolo_augmentation.py 11KB img_classification_augmentation.py 14KB __init__.py 888B __init__.py 960B img.py 3KB img_img.py 6KB triplet_net.py 7KB image_to_image nn.py 19KB __init__.py 730B losses yolo.py 13KB __init__.py 860B triplet_net triplet_mining.py 14KB triplet_loss.py 4KB __init__.py 892B generative_models images cgan.py 18KB gan.py 23KB __init__.py 865B cycle_gan.py 25KB __init__.py 729B __init__.py 1KB metric_learning triplet_utils.py 3KB knn_classifier.py 4KB __init__.py 891B similarity_attention.py 14KB triplet_net.py 36KB adversarial_attacks images __init__.py 754B fgsm.py 5KB __init__.py 733B layers __init__.py 831B yolo_decoder.py 10KB se_block.py 2KB non_local_block.py 6KB image_detection yolo.py 38KB __init__.py 735B interpretability __init__.py 737B grad_cam.py 9KB errors dnnlab_exceptions.py 822B __init__.py 0B utils coco_map.py 10KB performance.py 2KB gpu.py 1KB __init__.py 773B setup.cfg 38B setup.py 1KB README.md 771B资源评论
