A diverse collection of robotics, machine learning, computer vision, and data processing experiments and utilities. This repository serves as a personal workspace for exploring various technologies and implementing solutions across multiple domains.
This repository contains multiple independent projects covering:
- Robotics: LeRobot integration, forward/inverse kinematics, Isaac Sim teleoperation
- Machine Learning: Neural network implementations, optimizer comparisons, W&B experiments
- Computer Vision: Photo editing tools, MMDiT implementations
- Audio Processing: Speech-to-text transcription and AI summarization
- SAR Data Processing: Sentinel-1 data exploration and analysis
- Containerization: Docker environments for GPU-accelerated AI workflows
- Algorithms: LeetCode solutions and algorithmic practice
fun_project/
├── audio_transcribe/ # Audio transcription and AI summarization
├── basic_NN/ # Neural network experiments (Newton vs Adam optimizer)
├── docker/ # Docker configurations for development environments
│ └── isaac-brain/ # GPU-enabled Isaac Sim + IsaacLab environment
├── FK_IK/ # Forward and inverse kinematics utilities
├── lerobot/ # LeRobot dataset conversion and teleoperation scripts
├── leetcode/ # Algorithm practice and solutions
├── modern_robotics/ # Modern Robotics course implementations
├── photo_editor/ # Photo editing using Nano Banana (MMDiT)
├── sar/ # SAR (Synthetic Aperture Radar) data exploration
└── wandb/ # Weights & Biases experiment tracking utilities
Complete pipeline for transcribing audio files and generating AI summaries using local models:
- Speech-to-text with OpenAI Whisper (all model sizes)
- AI summarization with local LLMs (Ollama/llama.cpp)
- Multiple output formats (TXT, JSON, SRT, VTT)
- Batch processing support
- Privacy-first: All processing done locally
Quick start:
cd audio_transcribe
pip install -r requirements.txt
# Transcribe audio
python transcribe.py audio.mp3
# Summarize transcript
python summarize.py output/transcripts/audio.txt
# Or use the complete pipeline
python process_audio.py audio.mp3 --model base --summary-style meetingSee audio_transcribe/README.md for detailed instructions.
Comparison of optimization algorithms for training modified AlexNet on CIFAR-10:
- Custom Newton optimizer (second-order)
- Adam optimizer (first-order)
- Training metrics and performance analysis
Usage:
python basic_NN/main.pyScripts for working with LeRobot framework including:
- HDF5 dataset conversion from Isaac Sim to LeRobot format
- XLeRobot teleoperation setup and calibration
- Data upload utilities to Hugging Face
- Motor setup and calibration configurations
Key workflows:
- Motor setup:
lerobot-setup-motors - Calibration:
lerobot-calibrate - Teleoperation data collection with Isaac Sim
- Model inference with Gr00t and SmolVLA
See lerobot/notes.md for detailed instructions.
GPU-accelerated Docker container for AI model training with:
- NVIDIA CUDA 12.8 + cuDNN 9
- Python 3.11 with conda environment
- Isaac Sim 5.0 and IsaacLab 2.2.0 support
- Optimized for RTX 5080 (16GB)
Quick start:
cd docker/isaac-brain
docker-compose up -d
docker-compose exec isaac-brain bashSee docker/isaac-brain/README.md for complete setup instructions.
Workflow for downloading and analyzing Sentinel-1 SAR data:
- Query and download from ASF (Alaska Satellite Facility)
- Metadata extraction with pyroSAR
- Backscatter statistics and visualization
- Time series analysis potential
Features:
- VV-polarization processing
- Histogram generation
- Quick preview images
- Statistical analysis
See sar/README.md for more details.
Utilities for robotic kinematics calculations:
- Closed-loop control utilities
- Robot checkpoint management
- Integration with LeRobot/Isaac Sim workflows
Image editing tool using Nano Banana and MMDiT (Multimodal Diffusion Transformer):
- Custom MMDiT implementation
- Interactive editing interface
- Jupyter notebook for experimentation
Experiment tracking and model inference utilities:
- Multi-run experiment management
- Pi0 policy inference scripts
- Quick testing utilities
Core dependencies vary by project. See the following files for detailed requirements:
environment.yml: Conda environment specification (recommended)requirements_full.txt: Complete pip requirementsrequirement.txt: SAR-specific dependenciesSETUP.md: Detailed setup guide with troubleshooting
# Option 1: Conda (recommended)
conda env create -f environment.yml
conda activate fun_project
# Option 2: Pip
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
pip install -r requirements_full.txt
# Option 3: Quick setup script
./setup.sh conda # or 'pip' or 'docker'For detailed setup instructions, see SETUP.md.
This repository uses:
- Python: 3.11+
- CUDA: 12.8 (for GPU workloads)
- Docker: For containerized development
- Conda: For environment management
- Git: Version control
Current branch: main
Recent activity:
- LeRobot notes and configurations
- Docker setup for Isaac Sim/IsaacLab environments
- Deployment updates
- Recording management features
-
Clone the repository:
git clone <repository-url> cd fun_project
-
Choose your project: Navigate to the specific project directory and follow its README or inline documentation.
-
Set up dependencies: Install requirements specific to your chosen project.
-
Docker users: For GPU-accelerated workflows, see
docker/isaac-brain/for containerized environment setup.
- Audio Transcription: See audio_transcribe/README.md
- Neural Networks: See inline comments in
basic_NN/main.py - LeRobot: See lerobot/notes.md for detailed workflows
- Docker: See docker/isaac-brain/README.md
- SAR: See sar/README.md
- Photo Editor: See photo_editor/README.md
- Projects are mostly independent and can be used separately
- Some projects (LeRobot, FK_IK, wandb) may share dependencies
- Docker environment is optional but recommended for GPU workloads
- See individual project directories for specific usage instructions
This is a personal experimentation repository. Feel free to explore and adapt code for your own projects.
Not specified. Please contact the repository owner for licensing information.