diff options
| author | skal <pascal.massimino@gmail.com> | 2026-02-12 11:50:52 +0100 |
|---|---|---|
| committer | skal <pascal.massimino@gmail.com> | 2026-02-12 11:50:52 +0100 |
| commit | 7547e8ff4744339b92650b6ef3ff7405befe4beb (patch) | |
| tree | 0388b064c6bb2fcb2346796f9d1134c5ed9214b5 /scripts | |
| parent | c878631f24ddb7514dd4db3d7ace6a0a296d4157 (diff) | |
CNN v2: Patch-based training as default (like CNN v1)
Salient point detection on original images with patch extraction.
Changes:
- Added PatchDataset class (harris/fast/shi-tomasi/gradient detectors)
- Detects salient points on ORIGINAL images (no resize)
- Extracts 32×32 patches around salient points
- Default: 64 patches/image, harris detector
- Batch size: 16 (512 patches per batch)
Training modes:
1. Patch-based (default): --patch-size 32 --patches-per-image 64 --detector harris
2. Full-image (option): --full-image --image-size 256
Benefits:
- Focuses training on interesting regions
- Handles variable image sizes naturally
- Matches CNN v1 workflow
- Better convergence with limited data (8 images → 512 patches)
Script updated:
- train_cnn_v2_full.sh: Patch-based by default
- Configuration exposed for easy switching
Example:
./scripts/train_cnn_v2_full.sh # Patch-based
# Edit script: uncomment FULL_IMAGE for resize mode
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Diffstat (limited to 'scripts')
| -rwxr-xr-x | scripts/train_cnn_v2_full.sh | 20 |
1 files changed, 16 insertions, 4 deletions
diff --git a/scripts/train_cnn_v2_full.sh b/scripts/train_cnn_v2_full.sh index 119b788..4ddd9ac 100755 --- a/scripts/train_cnn_v2_full.sh +++ b/scripts/train_cnn_v2_full.sh @@ -14,8 +14,17 @@ CHECKPOINT_DIR="checkpoints" VALIDATION_DIR="validation_results" EPOCHS=10000 CHECKPOINT_EVERY=500 -BATCH_SIZE=8 -IMAGE_SIZE=256 +BATCH_SIZE=16 + +# Patch-based training (default) +PATCH_SIZE=32 +PATCHES_PER_IMAGE=64 +DETECTOR="harris" + +# Full-image training (alternative - uncomment to use) +# FULL_IMAGE="--full-image" +# IMAGE_SIZE=256 + KERNEL_SIZES="1 3 5" CHANNELS="16 8 4" @@ -31,13 +40,16 @@ echo "[1/4] Training CNN v2 model..." python3 training/train_cnn_v2.py \ --input "$INPUT_DIR" \ --target "$TARGET_DIR" \ - --image-size $IMAGE_SIZE \ + --patch-size $PATCH_SIZE \ + --patches-per-image $PATCHES_PER_IMAGE \ + --detector $DETECTOR \ --kernel-sizes $KERNEL_SIZES \ --channels $CHANNELS \ --epochs $EPOCHS \ --batch-size $BATCH_SIZE \ --checkpoint-dir "$CHECKPOINT_DIR" \ - --checkpoint-every $CHECKPOINT_EVERY + --checkpoint-every $CHECKPOINT_EVERY \ + $FULL_IMAGE if [ $? -ne 0 ]; then echo "Error: Training failed" |