deliberable_04_01_2026

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

src/README.md

@@ -95,6 +95,25 @@ Results are saved to `src/results/` as CSV files:
 - `raytune_doctr_results_<timestamp>.csv`
 - `raytune_easyocr_results_<timestamp>.csv`
 
+### Correlation Analysis
+
+Correlation tables used in the thesis are derived from the CSV results with a local script:
+
+```bash
+source .venv/bin/activate
+python tem/scripts/compute_correlations_all.py
+```
+
+Outputs are written to `src/results/correlations/`:
+- `paddle_correlations.csv`
+- `doctr_correlations.csv`
+- `easyocr_correlations.csv`
+
+These files are computed from the corresponding inputs:
+- `src/results/raytune_paddle_results_20260119_122609.csv`
+- `src/results/raytune_doctr_results_20260119_121445.csv`
+- `src/results/raytune_easyocr_results_20260119_120204.csv`
+
 ## Directory Structure
 
 ```

src/docker-compose.tuning.yml

@@ -1,82 +0,0 @@
-# docker-compose.tuning.yml - Ray Tune with all OCR services (PaddleOCR + DocTR)
-# Usage:
-#   docker compose -f docker-compose.tuning.yml up -d paddle-ocr-gpu doctr-gpu
-#   docker compose -f docker-compose.tuning.yml run raytune --service paddle --samples 64
-#   docker compose -f docker-compose.tuning.yml run raytune --service doctr --samples 64
-#   docker compose -f docker-compose.tuning.yml down
-#
-# Note: EasyOCR uses port 8002 (same as PaddleOCR). Use docker-compose.tuning.easyocr.yml separately.
-
-services:
-  raytune:
-    image: seryus.ddns.net/unir/raytune:latest
-    network_mode: host
-    shm_size: '5gb'
-    volumes:
-      - ./results:/app/results:rw
-    environment:
-      - PYTHONUNBUFFERED=1
-
-  paddle-ocr-gpu:
-    image: seryus.ddns.net/unir/paddle-ocr-gpu:latest
-    container_name: paddle-ocr-gpu-tuning
-    ports:
-      - "8002:8000"
-    volumes:
-      - ./dataset:/app/dataset:ro
-      - ./debugset:/app/debugset:rw
-      - paddlex-cache:/root/.paddlex
-    environment:
-      - PYTHONUNBUFFERED=1
-      - CUDA_VISIBLE_DEVICES=0
-      - PADDLE_DET_MODEL=PP-OCRv5_mobile_det
-      - PADDLE_REC_MODEL=PP-OCRv5_mobile_rec
-    deploy:
-      resources:
-        reservations:
-          devices:
-            - driver: nvidia
-              count: 1
-              capabilities: [gpu]
-    restart: unless-stopped
-    healthcheck:
-      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:8000/health')"]
-      interval: 30s
-      timeout: 10s
-      retries: 3
-      start_period: 60s
-
-  doctr-gpu:
-    image: seryus.ddns.net/unir/doctr-gpu:latest
-    container_name: doctr-gpu-tuning
-    ports:
-      - "8003:8000"
-    volumes:
-      - ./dataset:/app/dataset:ro
-      - ./debugset:/app/debugset:rw
-      - doctr-cache:/root/.cache/doctr
-    environment:
-      - PYTHONUNBUFFERED=1
-      - CUDA_VISIBLE_DEVICES=0
-      - DOCTR_DET_ARCH=db_resnet50
-      - DOCTR_RECO_ARCH=crnn_vgg16_bn
-    deploy:
-      resources:
-        reservations:
-          devices:
-            - driver: nvidia
-              count: 1
-              capabilities: [gpu]
-    restart: unless-stopped
-    healthcheck:
-      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:8000/health')"]
-      interval: 30s
-      timeout: 10s
-      retries: 3
-      start_period: 180s
-
-volumes:
-  paddlex-cache:
-    name: paddlex-model-cache
-  doctr-cache:
-    name: doctr-model-cache

src/results/correlations/doctr_correlations.csv

@@ -0,0 +1,19 @@
+parameter,metric,pearson
+straighten_pages,CER,0.9998131749398365
+symmetric_pad,CER,0.5261495908818205
+disable_page_orientation,CER,-0.49338806188002227
+disable_crop_orientation,CER,0.3470400052300961
+paragraph_break,CER,-0.28710190683729514
+resolve_blocks,CER,-0.2307325611083525
+preserve_aspect_ratio,CER,0.12408787013954682
+assume_straight_pages,CER,-0.11904725618816066
+resolve_lines,CER,0.060358529330885535
+straighten_pages,WER,0.9997423603542454
+symmetric_pad,WER,0.5288569133407047
+disable_page_orientation,WER,-0.49788185337361895
+disable_crop_orientation,WER,0.35150041517456027
+paragraph_break,WER,-0.2867204522845194
+resolve_blocks,WER,-0.23363320696739256
+assume_straight_pages,WER,-0.12980061688173894
+preserve_aspect_ratio,WER,0.1289984520450311
+resolve_lines,WER,0.06486504638919248

src/results/correlations/doctr_correlations.json

@@ -0,0 +1,92 @@
+[
+  {
+    "parameter": "straighten_pages",
+    "metric": "CER",
+    "pearson": 0.9998131749398365
+  },
+  {
+    "parameter": "symmetric_pad",
+    "metric": "CER",
+    "pearson": 0.5261495908818205
+  },
+  {
+    "parameter": "disable_page_orientation",
+    "metric": "CER",
+    "pearson": -0.49338806188002227
+  },
+  {
+    "parameter": "disable_crop_orientation",
+    "metric": "CER",
+    "pearson": 0.3470400052300961
+  },
+  {
+    "parameter": "paragraph_break",
+    "metric": "CER",
+    "pearson": -0.28710190683729514
+  },
+  {
+    "parameter": "resolve_blocks",
+    "metric": "CER",
+    "pearson": -0.2307325611083525
+  },
+  {
+    "parameter": "preserve_aspect_ratio",
+    "metric": "CER",
+    "pearson": 0.12408787013954682
+  },
+  {
+    "parameter": "assume_straight_pages",
+    "metric": "CER",
+    "pearson": -0.11904725618816066
+  },
+  {
+    "parameter": "resolve_lines",
+    "metric": "CER",
+    "pearson": 0.060358529330885535
+  },
+  {
+    "parameter": "straighten_pages",
+    "metric": "WER",
+    "pearson": 0.9997423603542454
+  },
+  {
+    "parameter": "symmetric_pad",
+    "metric": "WER",
+    "pearson": 0.5288569133407047
+  },
+  {
+    "parameter": "disable_page_orientation",
+    "metric": "WER",
+    "pearson": -0.49788185337361895
+  },
+  {
+    "parameter": "disable_crop_orientation",
+    "metric": "WER",
+    "pearson": 0.35150041517456027
+  },
+  {
+    "parameter": "paragraph_break",
+    "metric": "WER",
+    "pearson": -0.2867204522845194
+  },
+  {
+    "parameter": "resolve_blocks",
+    "metric": "WER",
+    "pearson": -0.23363320696739256
+  },
+  {
+    "parameter": "assume_straight_pages",
+    "metric": "WER",
+    "pearson": -0.12980061688173894
+  },
+  {
+    "parameter": "preserve_aspect_ratio",
+    "metric": "WER",
+    "pearson": 0.1289984520450311
+  },
+  {
+    "parameter": "resolve_lines",
+    "metric": "WER",
+    "pearson": 0.06486504638919248
+  }
+]

src/results/correlations/easyocr_correlations.csv

@@ -0,0 +1,25 @@
+parameter,metric,pearson
+contrast_ths,CER,0.40885606429688176
+ycenter_ths,CER,0.3052506223332593
+slope_ths,CER,-0.3007836023513022
+width_ths,CER,0.2820622927402215
+beamWidth,CER,0.24551397803020547
+add_margin,CER,0.23419721611930053
+height_ths,CER,-0.22762064209067434
+low_text,CER,-0.2127914870290739
+adjust_contrast,CER,-0.14938336246670267
+text_threshold,CER,-0.12089385052834749
+link_threshold,CER,-0.09553910777087017
+min_size,CER,0.04597421017746789
+contrast_ths,WER,0.38684315732406244
+slope_ths,WER,-0.32504213142289984
+ycenter_ths,WER,0.28738383141120705
+add_margin,WER,0.24148232685944232
+width_ths,WER,0.23869424685132606
+height_ths,WER,-0.23743923240967893
+beamWidth,WER,0.2286079838179018
+adjust_contrast,WER,-0.17842640649533945
+low_text,WER,-0.1772397092408802
+text_threshold,WER,-0.12733512825321042
+min_size,WER,0.10070615378426818
+link_threshold,WER,-0.04425190559911718

src/results/correlations/easyocr_correlations.json

@@ -0,0 +1,122 @@
+[
+  {
+    "parameter": "contrast_ths",
+    "metric": "CER",
+    "pearson": 0.40885606429688176
+  },
+  {
+    "parameter": "ycenter_ths",
+    "metric": "CER",
+    "pearson": 0.3052506223332593
+  },
+  {
+    "parameter": "slope_ths",
+    "metric": "CER",
+    "pearson": -0.3007836023513022
+  },
+  {
+    "parameter": "width_ths",
+    "metric": "CER",
+    "pearson": 0.2820622927402215
+  },
+  {
+    "parameter": "beamWidth",
+    "metric": "CER",
+    "pearson": 0.24551397803020547
+  },
+  {
+    "parameter": "add_margin",
+    "metric": "CER",
+    "pearson": 0.23419721611930053
+  },
+  {
+    "parameter": "height_ths",
+    "metric": "CER",
+    "pearson": -0.22762064209067434
+  },
+  {
+    "parameter": "low_text",
+    "metric": "CER",
+    "pearson": -0.2127914870290739
+  },
+  {
+    "parameter": "adjust_contrast",
+    "metric": "CER",
+    "pearson": -0.14938336246670267
+  },
+  {
+    "parameter": "text_threshold",
+    "metric": "CER",
+    "pearson": -0.12089385052834749
+  },
+  {
+    "parameter": "link_threshold",
+    "metric": "CER",
+    "pearson": -0.09553910777087017
+  },
+  {
+    "parameter": "min_size",
+    "metric": "CER",
+    "pearson": 0.04597421017746789
+  },
+  {
+    "parameter": "contrast_ths",
+    "metric": "WER",
+    "pearson": 0.38684315732406244
+  },
+  {
+    "parameter": "slope_ths",
+    "metric": "WER",
+    "pearson": -0.32504213142289984
+  },
+  {
+    "parameter": "ycenter_ths",
+    "metric": "WER",
+    "pearson": 0.28738383141120705
+  },
+  {
+    "parameter": "add_margin",
+    "metric": "WER",
+    "pearson": 0.24148232685944232
+  },
+  {
+    "parameter": "width_ths",
+    "metric": "WER",
+    "pearson": 0.23869424685132606
+  },
+  {
+    "parameter": "height_ths",
+    "metric": "WER",
+    "pearson": -0.23743923240967893
+  },
+  {
+    "parameter": "beamWidth",
+    "metric": "WER",
+    "pearson": 0.2286079838179018
+  },
+  {
+    "parameter": "adjust_contrast",
+    "metric": "WER",
+    "pearson": -0.17842640649533945
+  },
+  {
+    "parameter": "low_text",
+    "metric": "WER",
+    "pearson": -0.1772397092408802
+  },
+  {
+    "parameter": "text_threshold",
+    "metric": "WER",
+    "pearson": -0.12733512825321042
+  },
+  {
+    "parameter": "min_size",
+    "metric": "WER",
+    "pearson": 0.10070615378426818
+  },
+  {
+    "parameter": "link_threshold",
+    "metric": "WER",
+    "pearson": -0.04425190559911718
+  }
+]

src/results/correlations/paddle_correlations.csv

@@ -0,0 +1,15 @@
+parameter,metric,pearson
+use_doc_unwarping,CER,0.8791236551817551
+use_doc_orientation_classify,CER,-0.7119850615039771
+textline_orientation,CER,-0.5347452891182014
+text_det_thresh,CER,0.4280438958428758
+text_det_box_thresh,CER,0.3113152196833144
+text_rec_score_thresh,CER,-0.2681957118190106
+text_det_unclip_ratio,CER,
+use_doc_unwarping,WER,0.743651897463081
+use_doc_orientation_classify,WER,-0.6018981292243886
+textline_orientation,WER,-0.5906753653336065
+text_det_thresh,WER,0.39917807081409956
+text_det_box_thresh,WER,0.2555315418488065
+text_rec_score_thresh,WER,-0.08030912963602418
+text_det_unclip_ratio,WER,

src/results/correlations/paddle_correlations.json

@@ -0,0 +1,72 @@
+[
+  {
+    "parameter": "use_doc_unwarping",
+    "metric": "CER",
+    "pearson": 0.8791236551817551
+  },
+  {
+    "parameter": "use_doc_orientation_classify",
+    "metric": "CER",
+    "pearson": -0.7119850615039771
+  },
+  {
+    "parameter": "textline_orientation",
+    "metric": "CER",
+    "pearson": -0.5347452891182014
+  },
+  {
+    "parameter": "text_det_thresh",
+    "metric": "CER",
+    "pearson": 0.4280438958428758
+  },
+  {
+    "parameter": "text_det_box_thresh",
+    "metric": "CER",
+    "pearson": 0.3113152196833144
+  },
+  {
+    "parameter": "text_rec_score_thresh",
+    "metric": "CER",
+    "pearson": -0.2681957118190106
+  },
+  {
+    "parameter": "text_det_unclip_ratio",
+    "metric": "CER",
+    "pearson": NaN
+  },
+  {
+    "parameter": "use_doc_unwarping",
+    "metric": "WER",
+    "pearson": 0.743651897463081
+  },
+  {
+    "parameter": "use_doc_orientation_classify",
+    "metric": "WER",
+    "pearson": -0.6018981292243886
+  },
+  {
+    "parameter": "textline_orientation",
+    "metric": "WER",
+    "pearson": -0.5906753653336065
+  },
+  {
+    "parameter": "text_det_thresh",
+    "metric": "WER",
+    "pearson": 0.39917807081409956
+  },
+  {
+    "parameter": "text_det_box_thresh",
+    "metric": "WER",
+    "pearson": 0.2555315418488065
+  },
+  {
+    "parameter": "text_rec_score_thresh",
+    "metric": "WER",
+    "pearson": -0.08030912963602418
+  },
+  {
+    "parameter": "text_det_unclip_ratio",
+    "metric": "WER",
+    "pearson": NaN
+  }
+]