- Transfer learning
- YoloV4 configuration
- YoloV4 training
- YoloV4 loss function adjustments.
- Live plot losses
- Command line options
- YoloV3 tiny
- Rasberry Pi support
-
- (new) Command line options
- DarkNet models loaded directly from .cfg files
- YoloV4 support
- tensorflow-2.7--keras-functional-api
- cpu-gpu support
- Random weights and DarkNet weights support
- csv-xml annotation parsers.
- Anchor generator.
- matplotlib visualization of all stages.
- tf.data input pipeline.
- imgaug augmentation pipeline(customizable).
- Logging
- All-in-1 custom trainer.
- Stop and resume training support.
- Fully vectorized mAP evaluation.
- labelpix support.
- Photo & video detection
- Clone the repo
git clone https://github.com/schissmantics/yolo-tf2- Install
cd yolo-tf2
pip install .- Verify installation
yolotf2OUT:
Yolo-tf2 1.5
Usage:
yolotf2 <command> [options] [args]
Available commands:
train Create new or use existing dataset and train a model
evaluate Evaluate a trained model
detect Detect a folder of images or a video
yolo-tf2 was initially an implementation of yolov3 (you only look once)(training & inference) and support for all yolo versions was added in V1.1
Yolo's original repo is here (written in C/C++/Cu). Yolo is a state-of-the-art, real-time object detection system that is extremely fast and accurate. There are many implementations that support tensorflow, only a few that support tensorflow v2 and as I did not find versions that suit my needs so, I decided to create this version which is very flexible and customizable. It requires python 3.9+, is not platform specific and is MIT licensed.
- Fix bug that creates out of image bounding boxes after augmentation.
- Fix bug with display commands
- Update dependencies to most recent versions.
- Fix a bug that draws extra irrelevant boxes over photos for V4 configuration
- Fix a bug that causes shape incompatibility issues
- Remove extra required parameters
image_width,image_heightwhich are currently measured during runtime. - Fix a bug that prevents output from saving due to path mishandling
- Unify all IO operations to go through
yolo_tf2.utils.common.get_abs_path() - All commands are currently supported from any working directory, so users shouldn't worry about creating folders and matching names, which is handled automatically.
- Upgrade all dependencies to latest versions
- Add command line full support for training, evaluation and detection
- Add setup.py, direct installation through python3 setup.py install
- Restructure package folders
- Add yolov4 support
- Models are loaded directly from DarkNet .cfg files
- YoloV4 is currently supported(inference only, no training)
- Mish activation function(YoloV4)
| flags | help | required | default |
|---|---|---|---|
| --input-shape | Input shape ex: (416, 416, 3) | - | (416, 416, 3) |
| --classes | Path to classes .txt file | True | - |
| --model-cfg | Yolo DarkNet configuration .cfg file | True | - |
| --max-boxes | Maximum boxes per image | - | 100 |
| --iou-threshold | IOU(intersection over union threshold) | - | 0.5 |
| --score-threshold | Confidence score threshold | - | 0.5 |
| --workers | Concurrent tasks(in areas where that is possible) | - | 16 |
| --process-batch-size | Batch size of operations that needs batching (excluding training) | - | 32 |
| flags | help | required | default |
|---|---|---|---|
| --weights | Path to trained weights .tf or .weights file | - | - |
| --epochs | Number of training epochs | - | 100 |
| --batch-size | Training batch size | - | 8 |
| --learning-rate | Training learning rate | - | 0.001 |
| --dataset-name | Name of the checkpoint | True | - |
| --test-size | test dataset relative size (a value between 0 and 1) | - | 0.1 |
| --evaluate | If True, evaluation will be conducted after training | - | - |
| --merge-evaluation | If False, evaluate training and validation separately | - | - |
| --shuffle-buffer | Dataset shuffle buffer | - | 512 |
| --min-overlaps | a float value between 0 and 1 | - | 0.5 |
| --display-stats | If True, display evaluation statistics | - | - |
| --plot-stats | If True, plot results | - | - |
| --save-figs | If True, save plots | - | - |
| --clear-output | If True, clear output folders | - | - |
| --n-eval | Evaluate every n epochs | - | - |
| --relative-labels | Path to .csv file that contains | - | - |
| --voc-conf | VOC configuration .json file | - | - |
| --augmentation-preset | name of augmentation preset | - | - |
| --image-folder | Path to folder that contains images, defaults to data/photos | - | - |
| --xml-labels-folder | Path to folder that contains XML labels | - | - |
| --train-tfrecord | Path to training .tfrecord file | - | - |
| --valid-tfrecord | Path to validation .tfrecord file | - | - |
| flags | help | required |
|---|---|---|
| --predicted-data | csv file with predictions | True |
| --actual-data | csv file with actual data | True |
| --train-tfrecord | Path to training .tfrecord file | True |
| --valid-tfrecord | Path to validation .tfrecord file | True |
| flags | help | required | default |
|---|---|---|---|
| --image | Path to an image to predict and draw bounding boxes over | - | - |
| --image-dir | A directory that contains images to predict | - | - |
| --video | A video to predict | - | - |
| --codec | Codec to use for predicting videos | - | mp4v |
| --display-vid | Display video during prediction | - | - |
| --weights | Path to trained weights .tf or .weights file | True | - |
| --output-dir | Path to directory for saving results | - | - |
This feature was introduced to replace the old hard-coded model. Models are loaded directly from DarkNet .cfg files for convenience.
As models currently load from .cfg files directly, all yolo versions including v4 are supported the configuration file needs to be passed as argument, and the model is loaded.
This program leverages features that were introduced in tensorflow 2.x including:
- Eager execution: an imperative programming environment that evaluates operations immediately, without building graphs check here
tf.function: A JIT compilation decorator that speeds up some components of the program check heretf.data: API for input pipelines check here
The program detects and uses available GPUs at runtime(training/detection) if no GPUs available, the CPU will be used(slow).
Both options are available, and NOTE in case of using DarkNet yolov3 weights you must maintain the same number of COCO classes (80 classes) as transfer learning to models with different classes will be supported in future versions.
There are 2 currently supported formats that the program is able to read and translate to input.
- XML VOC format which looks like the following example:
<annotation>
<folder>/path/to/image/folder</folder>
<filename>image_filename.png</filename>
<path>/path/to/image/folder/image_filename.png</path>
<size>
<width>1344</width>
<height>756</height>
<depth>3</depth>
</size>
<object>
<name>Car</name>
<bndbox>
<xmin>873.0000007680001</xmin>
<ymin>402.0000001920001</ymin>
<xmax>1315.00000128</xmax>
<ymax>697.0000000320001</ymax>
</bndbox>
</object>
<object>
<name>Car</name>
<bndbox>
<xmin>550.999999872</xmin>
<ymin>404.999999838</ymin>
<xmax>883.000000512</xmax>
<ymax>711.000000018</ymax>
</bndbox>
</object>
<object>
<name>Car</name>
<bndbox>
<xmin>8.999999903999992</xmin>
<ymin>374.999999976</ymin>
<xmax>525.99999984</xmax>
<ymax>736.000000344</ymax>
</bndbox>
</object>
<object>
<name>Traffic Lights</name>
<bndbox>
<xmin>857.999999808</xmin>
<ymin>312.99999960599996</ymin>
<xmax>903.9999991679999</xmax>
<ymax>372.99999933</ymax>
</bndbox>
</object>
<object>
<name>Traffic Lights</name>
<bndbox>
<xmin>1220.99999952</xmin>
<ymin>91.999999854</ymin>
<xmax>1317.999999456</xmax>
<ymax>249.99999985799997</ymax>
</bndbox>
</object>
<object>
<name>Traffic Lights</name>
<bndbox>
<xmin>701.999999232</xmin>
<ymin>207.00000014399998</ymin>
<xmax>753.999998976</xmax>
<ymax>275.000000184</ymax>
</bndbox>
</object>
<object>
<name>Street Sign</name>
<bndbox>
<xmin>1220.99999952</xmin>
<ymin>91.999999854</ymin>
<xmax>1317.999999456</xmax>
<ymax>249.99999985799997</ymax>
</bndbox>
</object>
<object>
<name>Traffic Lights</name>
<bndbox>
<xmin>701.999999232</xmin>
<ymin>207.00000014399998</ymin>
<xmax>753.999998976</xmax>
<ymax>275.000000184</ymax>
</bndbox>
</object>
<name>Street Sign</name>
<bndbox>
<xmin>798.99999984</xmin>
<ymin>244.999999944</ymin>
<xmax>881.00000016</xmax>
<ymax>275.000000184</ymax>
</bndbox>
</annotation>- CSV with relative labels that looks like the following example:
| image | object_name | object_index | bx | by | bw | bh | # |
|---|---|---|---|---|---|---|---|
| img1.png | dog | 2 | 0.438616071 | 0.51521164 | 0.079613095 | 0.123015873 | |
| img1.png | car | 1 | 0.177827381 | 0.381613757 | 0.044642857 | 0.091269841 | |
| img2.png | Street Sign | 5 | 0.674107143 | 0.44047619 | 0.040178571 | 0.084656085 |
A k-means algorithm finds the optimal sizes and generates anchors with process visualization.
Including:
- k-means visualization:
- Generated anchors:
- Precision and recall curves:
- Evaluation bar charts:
- Actual vs. detections:
- Augmentation visualization:
Augmentation visualization (before/after) as shown below:
- Dataset pre/post augmentation view with bounding boxes:
You can always visualize different stages of the program using my other repo labelpix which is tool for drawing bounding boxes, but can also be used to visualize bounding boxes over images using csv files in the format mentioned here.
TFRecords a simple format for storing a sequence of binary records. Protocol buffers are a cross-platform, cross-language library for efficient serialization of structured data and are used as input pipeline to store and read data efficiently the program takes as input images and their respective annotations and builds training and validation(optional) TFRecords to be further used for all operations and TFRecords are also used in the evaluation(mid/post) training, so it's valid to say you can delete images to free space after conversion to TFRecords.
Special thanks to the amazing imgaug creators, an augmentation pipeline(optional) is available and NOTE that the augmentation is conducted before the training not during the training due to technical complications to integrate tensorflow and imgaug. If you have a small dataset, augmentation is an option and it can be preconfigured before the training.
Different operations are recorded using logging module.
For custom training, Trainer class accepts configurations for augmentation,
new anchor generation, new dataset(TFRecord(s)) creation, mAP evaluation
mid-training and post training. So all you have to do is place images
in data > photos, provide the configuration that suits you and start the training
process, all operations are managed from the same place for convenience.
For detailed instructions.
by default the trainer checkpoints to models > checkpoint_name.tf at the end of each training epoch which enables the training to be resumed at any given point by loading the checkpoint which would be the most recent.
Evaluation is optional during the training every n epochs(not recommended for large datasets as it predicts every image in the dataset) and one evaluation at the end which is optional as well. Training and validation datasets can be evaluated separately and calculate mAP(mean average precision) as well as precision and recall curves for every class in the model.
You can check my other repo labelpix which is a labeling tool for drawing bounding boxes over images if you need to make custom datasets the tool can help and is supported by the detector. You can use csv files in the format mentioned here as labels and load images if you need to preview any stage of the training/augmentation/evaluation/detection.
Detections can be performed on photos or videos using Detector class.
Here are the most basic steps to train using a custom dataset:
1- Create classes .txt file that contains classes delimited by \n
dog
cat
car
person
boat
fan
laptop
2- Create a training instance and specify input_shape, classes_file,
and either specify image_folder=/path/to/image/folder
or it would default to data > photos
from yolo_tf2.core.trainer import Trainer
trainer = Trainer(
input_shape=(416, 416, 3),
model_configuration='yolo_tf2/config/yolo3.cfg',
classes_file='/path/to/classes_file.txt',
image_folder=/path/to/image/folder
)
3- Create dataset configuration(dict) that contains the following keys:
dataset_name: TFRecord prefix(required)
and one of the following:(required)
relative_labels: path to csv file in the following format
or
xml_labels_folder: Path to folder containing xml labels (defaults todata > xml_labels)voc_conf: path to .json file containing VOC parsing configuration (you may use the one inyolotf2 > configor create a similar structure).
and
test_size: percentage of the validation split ex: 0.1(required)augmentation:True(optional)
and if augmentation this implies the following:
-
sequences: (required) A list of augmentation sequences. -
aug_workers: (optional) defaults to 32 parallel augmentations. -
aug_batch_size: (optional) this is the augmentation batch size defaults to 64 images to load at once.dataset_conf = { 'relative_labels': '/path/to/labels.csv', 'dataset_name': 'dataset_name', 'test_size': 0.2, 'sequences': PRESET1, # check Config > augmentation_options.py 'augmentation': True, }
4- Create new anchor generation configuration(dict) that contains the following keys(optional):
anchor_no: number of anchors(should be 9) and one of the following:-
relative_labels: same as dataset configuration above -
xml_labels_folder: same as dataset configuration aboveanchors_conf = { 'anchor_no': 9, 'relative_labels': '/path/to/labels.csv' } -
voc_conf: should be included ifxml_labels_folderis specified
-
5- Start the training
Note
If you're going to use DarkNet yolov3 weights, make sure the classes file contains 80 classes(COCO classes) or you'll get an error. Transfer learning to models with different number of classes will be supported in future versions of the program.
trainer.train(
epochs=100,
batch_size=8,
learning_rate=1e-3,
dataset_name='dataset_name',
merge_evaluation=False,
min_overlaps=0.5,
new_dataset_conf=dataset_conf, # check step 5
new_anchors_conf=anchors_conf, # check step 6
# weights='/path/to/weights' # If you're using DarkNet weights or resuming training
)
yolotf2 train --input-shape "(416, 416, 3)" --classes "path/to/classes.txt" --model-cfg "yolo_tf2/config/yolo3.cfg" --dataset-name "dataset_name" --relative-labels "path/to/labels.csv" --epochs 100 --batch-size 8 --learning-rate 1e3 --merge-evaluation --min-overlaps 0.5 --test-size 0.2 --augmentation-preset PRESET1 --image-folder /path/to/image/folder
Notes
- if you're training from outside the repo, specify --image-folder "your image folder"
- To train on an already existing dataset, specify --train-tfrecord and --valid-tfrecord
- You can specify to parse from xml folder directly using --xml-labels-folder if
outside the repo, otherwise, you might place labels in
data > xml_labels - You can specify weights using --weights
After the training completes:
- The trained model is saved in models folder(which you can use to resume training later/predict photos or videos)
- The resulting TFRecords and their corresponding csv data are saved in
data > tfrecords - The resulting figures and evaluation results are saved in output folder. And if you're training from outside the repo, the folders above will be created in the working directory(if they do not exist)
Here are the most basic steps to augment images(no training, just augmentation):
Note: Augmentation is supported through yolotf2 train --augmentation-preset PRESET1.
If you need to manually augment photos and take your time to examine/visualize the results, here are the steps:
1- Copy images to data > photos or specify image_folder=path_tom_image_folder
2- Ensure you have a csv file containing the labels in the format
mentioned here, if you have labels
in xml VOC format, you can easily convert them using utils > annotation_parsers.py parse_voc_folder()` (everything is explained in the docstrings)
3- Create augmentation instance:
from yolo_tf2.config.augmentation_options import augmentations
from yolo_tf2.core.augmentor import DataAugment
aug = DataAugment(
labels_file='/path/to/labels/csv/file',
augmentation_map=augmentations)
aug.create_sequences(sequences) # check the docs
aug.augment_photo_folder()
After augmentation you'll find augmented images in the data > photos folder or the folder you specified(if you did specify one)
And you should find 2 csv files in the output folder:
-
augmented_data_plus_original.csv: you can use this with labelpix to visualize results with bounding boxes -
adjusted_data_plus_original.csv
and any of the 2 csv files above can be used in the new dataset configuration in the training.
Here are the most basic steps to evaluate a trained model:
-
Create an evaluation instance:
from yolo_tf2.core.evaluator import Evaluator evaluator = Evaluator( input_shape=(416, 416, 3), model_configuration='yolo_tf2/config/yolo3.cfg', train_tf_record='/path/to/train.tfrecord', valid_tf_record='/path/to/valid.tfrecord', classes_file='/path/to/classes.txt', anchors=anchors, # defaults to yolov3 anchors score_threshold=0.1 # defaults to 0.5 but it's okay to be lower ) -
Read actual and prediction results(that resulted from the training)
actual = pd.read_csv('data/tfrecords/full_data.csv') preds = pd.read_csv('output/full_dataset_predictions.csv') -
Calculate mAP(mean average precision):
evaluator.calculate_map( prediction_data=preds, actual_data=actual, min_overlaps=0.5, display_stats=True)
yolotf2 evaluate --input-shape "(416, 416, 3)" --model-cfg "yolo_tf2/config/yolo3.cfg" --train-tfrecord "/path/to/train.tfrecord" --valid-tfrecord "/path/to/valid.tfrecord" --score-threshold 0.1 --predicted-data "output/full_dataset_predictions.csv" --actual-data "data/tfrecords/full_data.csv"
After evaluation, you'll find resulting plots and predictions in the output folder.
Here are the most basic steps to perform detection:
-
Create a detection instance:
from yolo_tf2.core.detector import Detector detector = Detector( input_shape=(416, 416, 3), model_configuration='/path/to/DarkNet/yolo_version.cfg, classes_file='/path/to/classes_file.txt', score_threshold=0.5, iou_threshold=0.5, max_boxes=100, anchors=anchors # Optional if not specified, yolo default anchors are used ) -
Perform detections:
A) Photos:
photos = ['photo/path1', 'photo/path2']
detector.predict_photos(photos=photos,
trained_weights='/path/to/trained/weights') # .tf or yolov3.weights(80 classes)
yolotf2 detect --input-shape "(416, 416, 3)" --classes "path/to/classes.txt" --model-cfg "yolo_tf2/config/yolo3.cfg" --score-threshold 0.5 --iou-threshold 0.5 --image-dir "path/to/image/dir" --weights "path/to/weights"
or alternatively, if you want to perform detection on a single image specify --image instead of --image-dir
B) Video
detector.detect_video(
'/path/to/target/vid',
'/path/to/trained/weights.tf',
)
yolotf2 detect --input-shape "(416, 416, 3)" --classes "path/to/classes.txt" --model-cfg "yolo_tf2/config/yolo3.cfg" --score-threshold 0.5 --iou-threshold 0.5 --video "path/to/video" --weights "path/to/weights"
After predictions is complete you'll find photos/video in output > detections
Contributions are what make the open source community such an amazing place to
learn, inspire, and create. Any contributions you make are greatly appreciated.
- Fork the Project
- Create your Feature Branch (
git checkout -b feature/AmazingFeature) - Commit your Changes (
git commit -m 'Add some AmazingFeature') - Push to the Branch (
git push origin feature/AmazingFeature) - Open a Pull Request
There are relevant cases in which the issues will be addressed and irrelevant ones that will be closed.
The following issues will be addressed.
- Bugs.
- Performance issues.
- Installation issues.
- Documentation issues.
- Feature requests.
- Dependency issues that can be solved.
The following issues will not be addressed and will be closed.
- Issues without context / clear and concise explanation.
- Issues without standalone code (minimum reproducible example), or a jupyter notebook link to reproduce errors.
- Issues that are improperly formatted.
- Issues that are dataset / label specific without a dataset sample link.
- Issues that are the result of doing something that is unsupported by the existing features.
- Issues that are not considered as improvement / useful.
Distributed under the MIT License. See LICENSE for more information.
Give a ⭐️ if this project helped you!
Project link: https://github.com/schissmantics/yolo-tf2