https://github.com/bigmb/mb_pytorch
malav pytorch files
https://github.com/bigmb/mb_pytorch
Last synced: 8 days ago
JSON representation
malav pytorch files
- Host: GitHub
- URL: https://github.com/bigmb/mb_pytorch
- Owner: bigmb
- Created: 2023-02-18T19:23:05.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2024-05-23T03:20:59.000Z (12 months ago)
- Last Synced: 2024-05-23T03:45:36.593Z (12 months ago)
- Language: Python
- Size: 506 KB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# MB PyTorch
A comprehensive PyTorch-based machine learning library providing implementations for object detection, classification, segmentation, and meta-learning tasks.
[](https://github.com/bigmb/mb_pytorch/actions/workflows/stale.yml)
[](https://pepy.tech/project/mb-pytorch)
## Features
- **Object Detection**: Implementation of object detection models and training pipelines
- **Classification**: Tools and models for image classification tasks
- **Segmentation**: Semantic segmentation models and utilities
- **Meta Learning**: Support for meta-learning approaches
- **Flexible Data Loading**: Configurable data loading with YAML-based configuration
- **Visualization**: Integration with Gradio and TensorBoard for model visualization
## Installation
Requirements:
- Python >= 3.8
- PyTorch
- torchvision
```bash
pip install mb_pytorch
```
Additional dependencies:
```
torchsummary
numpy
matplotlib
mb_pandas
mb_utils
onnx
onnx2pytorch
seaborn
tqdm
pytorch_grad_cam
pillow
opencv-python
mb_base
```
## Usage
### Object Detection Example
```python
from mb_pytorch.utils.yaml_reader import YAMLReader
from mb_pytorch.dataloader.loader import BaseDataset, TorchDataLoader
from mb_pytorch.detection.training import DetectionTrainer
# Load configuration
yaml_read = YAMLReader('scripts/detection/object_detection.yaml')
yaml_data = yaml_read.read()
# Create datasets
train_dataset = BaseDataset(
data_config=yaml_data['data']['file'],
task_type=yaml_data['model']['model_type'],
transform=yaml_data['transformation'],
is_train=True
)
val_dataset = BaseDataset(
data_config=yaml_data['data']['file'],
task_type=yaml_data['model']['model_type'],
transform=yaml_data['transformation'],
is_train=False
)
# Create dataloaders
train_dataloader = TorchDataLoader(
dataset=train_dataset,
batch_size=yaml_data['train_params']['batch_size'],
shuffle=yaml_data['train_params']['shuffle'],
num_workers=yaml_data['train_params']['num_workers']
)
val_dataloader = TorchDataLoader(
dataset=val_dataset,
batch_size=yaml_data['val_params']['batch_size'],
shuffle=yaml_data['val_params']['shuffle'],
num_workers=yaml_data['val_params']['num_workers']
)
# Initialize trainer and train
trainer = DetectionTrainer(
yaml_data,
train_dataloader,
val_dataloader,
device='cpu',
use_all_cpu_cores=True
)
trainer.train()
```
### Data Loading
The library provides flexible data loading capabilities through YAML configuration:
```python
from mb_pytorch.dataloader.loader import DataLoader
from mb_utils.src.logging import logger
loader = DataLoader('./scripts/loader_y.yaml')
out1, out2, o1, o2 = loader.data_load(logger=logger)
```
## Project Structure
```
mb_pytorch/
├── classification/ # Classification models and utilities
├── dataloader/ # Data loading utilities
├── detection/ # Object detection implementations
├── metalearning/ # Meta-learning implementations
├── metrics/ # Evaluation metrics
├── models/ # Base model implementations
├── segmentation/ # Segmentation models
├── training/ # Training utilities
└── utils/ # Helper utilities
```
## Visualization
The library supports multiple visualization options:
- **Gradio**: Interactive model visualization and testing
- **TensorBoard**: Training metrics and model performance visualization
## Examples
Check the `examples/` directory for Jupyter notebooks demonstrating various use cases:
- Object Detection
- Basic Data Loading
- Embedding Data Loading
- Model Visualization
- Segmentation Models