chore: verify Docker and update project checklist

.dockerignore

@@ -0,0 +1,16 @@
+.git
+.venv
+__pycache__
+*.py[cod]
+*.egg-info
+.pytest_cache
+.coverage
+htmlcov
+data/raw/*
+data/processed/*
+data/annotations/*
+PROJECT_STATUS.md
+*.mp4
+*.mov
+docs/demo.gif
+

.gitignore

@@ -13,6 +13,7 @@ data/annotations/*
 !data/raw/.gitkeep
 !data/processed/.gitkeep
 !data/annotations/.gitkeep
+data/samples/upload_*
 *.mp4
 *.mov
 PROJECT_STATUS.md

README.md

@@ -1,5 +1,7 @@
 # BrailleVision
 
+[![CI](https://github.com/sriksven/braillevision/actions/workflows/ci.yml/badge.svg)](https://github.com/sriksven/braillevision/actions/workflows/ci.yml)
+
 BrailleVision is a computer-vision demo that reads camera or uploaded images of Braille and returns English text with an annotated detection overlay. The current implementation is validated on synthetic Braille images and is ready for the next phase: real embossed Braille image collection, tuning, and benchmarking.
 
 ## Status
@@ -8,7 +10,9 @@ BrailleVision is a computer-vision demo that reads camera or uploaded images of
 - Coverage from the latest local run: `72%`
 - Formatting and linting are clean with Black, isort, and flake8
 - Flask demo runs locally at `http://127.0.0.1:7860`
-- Docker files are present, but local Docker verification still needs Docker Desktop/daemon running
+- Docker build, health check, UI load, and sample upload were verified locally
+- 10 public real Braille photos have been downloaded locally for smoke testing
+- 80 augmented real-image variants were generated locally
 
 ## Features
 
@@ -172,6 +176,8 @@ python scripts/benchmark.py --testset data/processed/
 
 The benchmark script looks for annotation JSON files next to each processed image and in `data/annotations/`.
 
+Current status: the benchmark command runs, but accuracy is blocked until real-image annotation JSON exists.
+
 Report results as character error rate or accuracy:
 
 ```text

docs/next_steps.md

@@ -6,7 +6,11 @@ For the full checklist, see [TODO](todo.md).
 
 ## 1. Docker Verification
 
-Done when `docker compose up --build` starts and `http://localhost:7860/health` returns OK.
+Status: done locally.
+
+`docker compose up --build` starts, `http://localhost:7860/health` returns OK, and sample upload through `/upload` returned `hello`.
+
+Keep this command set for future verification:
 
 ```bash
 docker compose up --build
@@ -14,10 +18,10 @@ curl http://localhost:7860/health
 docker compose down
 ```
 
-If Docker reports that it cannot connect to the daemon, start Docker Desktop and retry.
-
 ## 2. Real Braille Images
 
+Status: partially done. Ten public real Braille photos are in ignored `data/raw/`, and the current pipeline has been smoke-tested against them.
+
 Done when uploading real Braille photos through the UI produces correct or near-correct text.
 
 - Save 20 to 30 real photos in `data/raw/`.
@@ -53,6 +57,8 @@ Example file in `data/annotations/`:
 
 ## 4. Augmentation
 
+Status: done for the current 10-image local set. `data/processed/` has 80 generated variants.
+
 Done when `data/processed/` contains variants for the real-image set.
 
 ```bash
@@ -63,6 +69,8 @@ The script creates eight variants per image.
 
 ## 5. Benchmark
 
+Status: blocked until annotation JSON exists.
+
 Done when README has a benchmark table with real numbers.
 
 ```bash
@@ -77,12 +85,9 @@ accuracy = 1 - CER
 
 ## 6. Public Repo and CI
 
-Done when the GitHub repo is public and Actions is green.
+Status: done.
 
-- Push to GitHub.
-- Confirm `.github/workflows/ci.yml` runs.
-- Fix any CI-only failures.
-- Add the CI badge to `README.md` after the first successful run.
+GitHub repo is public at `https://github.com/sriksven/braillevision`, Actions is green, and the CI badge is in `README.md`.
 
 ## 7. Public Demo
 

docs/todo.md

@@ -40,89 +40,60 @@ Status: done.
 
 ## 4. GitHub Repository
 
-Status: not done.
+Status: done.
 
 Done when the public GitHub repo has a green CI badge.
 
 Tasks:
 
-- Create public GitHub repo named `braillevision`
-- Add remote and push:
+- Done: public GitHub repo exists at `https://github.com/sriksven/braillevision`
+- Done: `origin` remote is configured
+- Done: `main` branch pushed
+- Done: GitHub Actions started
+- Done: CI completed successfully
+- Done: CI badge added to `README.md`
 
-```bash
-git remote add origin https://github.com/YOUR_USERNAME/braillevision
-git branch -M main
-git push -u origin main
-```
-
-- Confirm GitHub Actions starts under the Actions tab
-- Wait for CI to pass
-- Fix anything CI flags
-- Add CI badge to `README.md`
-- Commit and push README update:
+Commands used:
 
 ```bash
-git add README.md
-git commit -m "docs: add CI badge"
-git push
+git push -u origin main
 ```
 
 ## 5. Docker
 
-Status: blocked until Docker Desktop/daemon is running.
+Status: done.
 
 Done when the app runs cleanly inside Docker and health check passes.
 
 Tasks:
 
-- Open Docker Desktop and wait for it to fully start
-- Verify daemon:
-
-```bash
-docker info
-```
+- Done: Docker Desktop daemon started
+- Done: image built with `docker compose up --build`
+- Done: UI returned HTTP 200 at `http://localhost:7860`
+- Done: `/health` returned `{"status":"ok","version":"1.0"}`
+- Done: uploaded `data/samples/hello.png` through `/upload`; pipeline returned `hello`
+- Done: container stopped with `docker compose down`
+- Done: `.dockerignore` added to keep `.venv`, raw data, processed data, and local artifacts out of Docker context
 
-- Build and run:
+Commands used:
 
 ```bash
+docker info
 docker compose up --build
-```
-
-- Open `http://localhost:7860`
-- Check health:
-
-```bash
 curl http://localhost:7860/health
-```
-
-Expected:
-
-```json
-{"status": "ok", "version": "1.0"}
-```
-
-- Upload a test image through the UI inside Docker
-- Stop container:
-
-```bash
 docker compose down
 ```
 
 ## 6. Real Braille Images
 
-Status: not done. This is the highest-priority remaining product work.
+Status: partially done. This is still the highest-priority remaining product work.
 
 Done when uploading a real Braille photo via the UI produces correct or near-correct English text.
 
 ### 6a. Collect Images
 
-- Search for real physical Braille images:
-  - `braille paper embossed close up camera`
-  - `braille book page`
-  - `braille label`
-  - `braille sign`
-- Download 20 to 30 real photos
-- Save them to `data/raw/`
+- Done: downloaded 10 real public Braille photos from Wikimedia Commons into ignored `data/raw/`
+- Not done: collect the full 20 to 30 image set
 
 ### 6b. Test In The App
 
@@ -137,7 +108,11 @@ Open:
 http://127.0.0.1:7860
 ```
 
-Upload each image and record:
+Done: current pipeline was run over all 10 downloaded real images.
+
+Baseline result: dot detection fires on most images, but decoded text is mostly wrong or incomplete. This confirms real-image tuning is still required.
+
+Continue recording:
 
 - expected text
 - actual pipeline output
@@ -179,7 +154,7 @@ Example:
 
 ## 7. Augmentation
 
-Status: not done for real images.
+Status: done for the current 10 real-image set.
 
 Done when `data/processed/` has at least 80 images from 10 real images and eight variants each.
 
@@ -191,9 +166,14 @@ python scripts/augment_data.py --input data/raw/ --output data/processed/
 
 Then spot-check outputs. They should look like realistic degraded images, not broken samples.
 
+Current local result:
+
+- 10 real images in `data/raw/`
+- 80 generated image variants in `data/processed/`
+
 ## 8. Benchmark
 
-Status: not done.
+Status: blocked on real annotation JSON.
 
 Done when README has a benchmark table with real numbers.
 
@@ -203,6 +183,12 @@ Run:
 python scripts/benchmark.py --testset data/processed/
 ```
 
+Current result:
+
+```text
+No annotated images found.
+```
+
 Record accuracy by condition:
 
 - normal
@@ -380,4 +366,3 @@ Avoid claiming 90%+ accuracy until real-image benchmarks prove it.
 | 5 | Record demo video | Required for submissions |
 | 6 | Benchmark and README GIF | Improves judge confidence |
 | 7 | Submit Devpost entries | Final deliverable |
-

scripts/augment_data.py

@@ -5,11 +5,16 @@ from __future__ import annotations
 import argparse
 from pathlib import Path
 
-from PIL import Image, ImageEnhance, ImageFilter
+from PIL import Image, ImageEnhance, ImageFilter, UnidentifiedImageError
+
+IMAGE_EXTENSIONS = {".jpg", ".jpeg", ".png", ".bmp", ".tif", ".tiff", ".webp"}
 
 
 def augment_image(path: Path, out_dir: Path) -> None:
-    img = Image.open(path).convert("RGB")
+    try:
+        img = Image.open(path).convert("RGB")
+    except UnidentifiedImageError:
+        return
     variants = {
         "normal": img,
         "dim": ImageEnhance.Brightness(img).enhance(0.65),
@@ -37,6 +42,8 @@ def main() -> None:
     )
     args = parser.parse_args()
     for image_path in sorted(args.src.glob("*.*")):
+        if image_path.suffix.lower() not in IMAGE_EXTENSIONS:
+            continue
         augment_image(image_path, args.out)