nimslo auto-aligning gif processor

what it does

• processes nimslo 3d film shots into smooth animated gifs
• uses cnn-based border detection for precise alignment
• applies histogram matching for consistent exposure
• creates bounce-effect gifs with automatic quality optimization

features

• manual image selection - pick and order your best 4-6 frames
• cnn alignment - sophisticated border detection and image subtraction
• automatic cropping - removes black bars from aligned images
• quality settings - high, medium, or optimized output
• bounce effect - smooth back-and-forth animation
• streamlined workflow - no confirmation dialogs, just results

quick start

# clone and setup
git clone <repo-url>
cd nimslo_processing

# create environment and alias
conda env create -f environment.yml
echo 'alias nimslo="cd $(pwd) && ./run_nimslo.sh"' >> ~/.zshrc
source ~/.zshrc

workflow

1. run nimslo - launches the processor
2. select folder - choose your nimslo image directory
3. pick images - use the gui to select and order 4-6 frames
4. crop area - drag to select the region to align
5. add reference points - double-click to mark alignment points (optional but recommended)
6. automatic processing - intelligent alignment, histogram matching, gif creation
7. output - find your gif in nimslo_gifs/

output

• gifs saved to nimslo_gifs/nimslo_high_TIMESTAMP.gif
• previews saved as preview_original.png, preview_aligned.png, preview_final.png
• quality levels indicated in filename

technical details

• intelligent alignment - reference point-based (photoshop-style) or cnn border detection
• sift feature matching - robust point correspondence detection
• multi-transformation support - homography (3+ points) or affine (2 points)
• quality validation - ssim, mse, and point error metrics for alignment assessment
• fallback system - graceful degradation when point matching fails
• automatic cropping - removes transformation artifacts
• bounce animation - forward + backward sequence for smooth motion

dependencies

• tensorflow 2.16+ (for cnn alignment)
• opencv-python (computer vision + sift features)
• pillow (image processing)
• matplotlib (previews)
• tkinter (gui)
• scikit-image (quality metrics)

all managed via conda environment - just run conda env create -f environment.yml