Deep Stream Integration for Face Conversion
Budget: $1,500 – $3,000 USD
I need an expert in NVIDIA Jetson and Deep Stream to help me integrate several components of my real-time face conversion project.
Current Pipeline:
- Input from RTSP
- Face detection and tracking
- Face embedding information extraction
- Face attribute information extraction
- Virtual face generation
- Face swap
- Virtual face conversion
- Output to RTSP
I've already converted the face detection, tracking, and embedding information extraction to Deep Stream. Now, I need to:
1. Convert the face swap model (currently using the open-source insight face from https://github.com/deepinsight/insightface) to Deep Stream.
2. Modify the virtual face generation to query the extracted attribute information using ChromaDB.
3. Convert the face attribute extraction model (solider) to Deep Stream.
Key Requirements:
- The main objective for converting these models is to ensure better integration with the existing pipeline.
- The face swap model is open-source.
- The speed of the face attribute extraction model is crucial for the project's success.
Ideal skills for the job include experience with Deep Stream, NVIDIA Jetson, and working with face swap models, particularly open-source ones. Speed and efficiency are key, so a background in high-performance computing or related fields would be beneficial.
Current Pipeline:
- Input from RTSP
- Face detection and tracking
- Face embedding information extraction
- Face attribute information extraction
- Virtual face generation
- Face swap
- Virtual face conversion
- Output to RTSP
I've already converted the face detection, tracking, and embedding information extraction to Deep Stream. Now, I need to:
1. Convert the face swap model (currently using the open-source insight face from https://github.com/deepinsight/insightface) to Deep Stream.
2. Modify the virtual face generation to query the extracted attribute information using ChromaDB.
3. Convert the face attribute extraction model (solider) to Deep Stream.
Key Requirements:
- The main objective for converting these models is to ensure better integration with the existing pipeline.
- The face swap model is open-source.
- The speed of the face attribute extraction model is crucial for the project's success.
Ideal skills for the job include experience with Deep Stream, NVIDIA Jetson, and working with face swap models, particularly open-source ones. Speed and efficiency are key, so a background in high-performance computing or related fields would be beneficial.