video.mp4
This application allows users to load a video, control playback (frame-by-frame, scrubbing, play/pause), and export individual frames as images.
- Python 3.7 or later
- Native video codecs installed so the QT Media Player can decode/playback the video and extract frames: https://www.codecguide.com/download_kl.htm (the basic package is likely enough, pay attention during install to avoid extra software)
- PyQt5 (installed from requirements.txt below)
-
Clone the repository:
git clone https://github.com/yourusername/frame_extractor.git cd frame_extractor -
Create a virtual environment and activate it:
python -m venv venv source venv/bin/activate # Linux/macOS venv\Scripts\activate # Windows -
Install the required dependencies:
pip install -r requirements.txt
-
Run the application:
python main.py -
Load a video file by clicking the "Open" button.
-
Control the video playback using the play/pause button, slider, and next frame button.
-
Click the "Export Frame" button to save the current frame as an image.
-
The exported frames will be saved in the "output" folder within the project directory.
To use the output from this program as input for colmap and subsequent instant-ngp processing assuming you have instant-ngp installed you can run
python C:\Users\shaun\Development\instant-ngp\scripts\colmap2nerf.py --images ./output --run_colmap --aabb_scale 64
This will generate a transforms.json file in the current working directory that you can drag and drop onto the instant-ngp UI (launch separately) to start training of a NeRF model.
Once the model is generated you can create a set of camera paths within instant-ngp, save out the camera positions as JSON, and save the ingp file (model snapshot) to disk and then run a command similar to the following to process the model and JSON into a video output as seen above:
py -3.9 C:\Users\shaun\Development\instant-ngp\scripts\run.py --load_snapshot=construction.ingp --video_camera_path=construction2_cam.json --video_n_seconds=10