feat: replace Python seam carving with caire Go binary

Replace the Python seam-carving library with caire (esimov/caire v1.5.0),
a Go-based content-aware resize engine that is faster and supports both
shrinking and enlarging via seam insertion.

- Add Go builder stage in Dockerfile to compile caire from source
- Rewrite seam-carving.ts to call caire via execFile (no Python sidecar)
- Remove content-aware-resize from PYTHON_SIDECAR_TOOLS (60s timeout)
- Add new options: blur radius, edge sensitivity, square mode, face detection
- Move content-aware toggle below standard resize in UI (subtler placement)
- Rename "Don't enlarge" to "Limit to original size" with hover tooltip
- Add smooth progress bar for medium-duration tools
- Delete seam_carve.py and remove seam-carving pip dependency
- Update integration tests and visual regression screenshots
This commit is contained in:
Siddharth Kumar Sah
2026-04-11 17:49:28 +08:00
parent b8227c45b9
commit 1707521f3a
13 changed files with 466 additions and 327 deletions
-1
View File
@@ -7,4 +7,3 @@ onnxruntime-gpu==1.20.1
numpy==1.26.4
Pillow==11.1.0
opencv-python-headless==4.10.0.84
seam-carving==1.1.0
-1
View File
@@ -7,4 +7,3 @@ onnxruntime==1.20.1
numpy==1.26.4
Pillow==11.1.0
opencv-python-headless==4.10.0.84
seam-carving==1.1.0
-161
View File
@@ -1,161 +0,0 @@
"""
Content-aware image resize using seam carving.
Uses the seam-carving library (li-plus) with optional face protection via MediaPipe.
Args:
sys.argv[1]: input image path
sys.argv[2]: output image path
sys.argv[3]: JSON settings string with keys:
- width (int, optional): target width
- height (int, optional): target height
- protectFaces (bool, optional): enable face detection for protection mask
"""
import json
import sys
def emit_progress(percent, stage):
"""Emit structured progress to stderr for bridge.ts to capture."""
print(json.dumps({"progress": percent, "stage": stage}), file=sys.stderr, flush=True)
def build_face_mask(img_array):
"""Detect faces with MediaPipe and return a boolean keep_mask."""
import numpy as np
try:
import mediapipe as mp
except ImportError:
emit_progress(20, "MediaPipe not available, skipping face protection")
return None
h, w = img_array.shape[:2]
mask = np.zeros((h, w), dtype=bool)
face_detection = mp.solutions.face_detection
detector = face_detection.FaceDetection(model_selection=1, min_detection_confidence=0.5)
try:
results = detector.process(img_array)
if not results.detections:
emit_progress(20, "No faces detected")
return None
for detection in results.detections:
bbox = detection.location_data.relative_bounding_box
x = int(bbox.xmin * w)
y = int(bbox.ymin * h)
bw = int(bbox.width * w)
bh = int(bbox.height * h)
# Add 20% padding around face
pad_x = int(bw * 0.2)
pad_y = int(bh * 0.2)
x1 = max(0, x - pad_x)
y1 = max(0, y - pad_y)
x2 = min(w, x + bw + pad_x)
y2 = min(h, y + bh + pad_y)
mask[y1:y2, x1:x2] = True
emit_progress(20, f"Detected {len(results.detections)} face(s)")
return mask
finally:
detector.close()
def main():
if len(sys.argv) < 4:
print(json.dumps({"success": False, "error": "Usage: seam_carve.py <input> <output> <settings>"}))
sys.exit(1)
input_path = sys.argv[1]
output_path = sys.argv[2]
try:
settings = json.loads(sys.argv[3])
except (json.JSONDecodeError, ValueError):
print(json.dumps({"success": False, "error": "Invalid settings JSON"}))
sys.exit(1)
target_width = settings.get("width")
target_height = settings.get("height")
protect_faces = settings.get("protectFaces", False)
try:
import numpy as np
from PIL import Image
except ImportError:
print(json.dumps({"success": False, "error": "Pillow/numpy not installed"}))
sys.exit(1)
try:
import seam_carving
except ImportError:
print(json.dumps({"success": False, "error": "seam-carving package not installed"}))
sys.exit(1)
try:
emit_progress(0, "Loading image")
img = Image.open(input_path).convert("RGB")
img_array = np.array(img)
src_h, src_w = img_array.shape[:2]
# Default to source dimensions if not specified
if target_width is None:
target_width = src_w
if target_height is None:
target_height = src_h
# Validate: shrink only
if target_width > src_w or target_height > src_h:
print(json.dumps({
"success": False,
"error": f"Content-aware resize only supports shrinking. Source is {src_w}x{src_h}, target is {target_width}x{target_height}."
}))
sys.exit(1)
# Nothing to do
if target_width == src_w and target_height == src_h:
img.save(output_path)
print(json.dumps({"success": True, "width": src_w, "height": src_h}))
return
# Warn about large images
if src_w > 3000 or src_h > 3000:
emit_progress(5, "Large image detected, this may take a while")
# Face protection mask
keep_mask = None
if protect_faces:
emit_progress(10, "Detecting faces")
keep_mask = build_face_mask(img_array)
emit_progress(25, "Starting seam carving")
# seam_carving.resize takes size as (width, height)
result = seam_carving.resize(
img_array,
(target_width, target_height),
energy_mode="backward",
order="width-first",
keep_mask=keep_mask,
)
emit_progress(90, "Saving result")
Image.fromarray(result).save(output_path)
print(json.dumps({
"success": True,
"width": result.shape[1],
"height": result.shape[0],
}))
except Exception as e:
print(json.dumps({"success": False, "error": str(e)}))
sys.exit(1)
if __name__ == "__main__":
main()
+75 -21
View File
@@ -1,11 +1,19 @@
import { readFile, writeFile } from "node:fs/promises";
import { execFile } from "node:child_process";
import { randomUUID } from "node:crypto";
import { readFile, rm, writeFile } from "node:fs/promises";
import { join } from "node:path";
import { type ProgressCallback, runPythonWithProgress } from "./bridge.js";
import { promisify } from "node:util";
import sharp from "sharp";
const execFileAsync = promisify(execFile);
export interface SeamCarveOptions {
width?: number;
height?: number;
protectFaces?: boolean;
blurRadius?: number;
sobelThreshold?: number;
square?: boolean;
}
export interface SeamCarveResult {
@@ -14,31 +22,77 @@ export interface SeamCarveResult {
height: number;
}
/**
* Discover the caire binary. Checks PATH (Docker installs to /usr/local/bin)
* and the CAIRE_PATH env var for local development.
*/
let cachedCairePath: string | null = null;
async function findCaire(): Promise<string> {
if (cachedCairePath) return cachedCairePath;
const candidates = process.env.CAIRE_PATH ? [process.env.CAIRE_PATH, "caire"] : ["caire"];
for (const cmd of candidates) {
try {
await execFileAsync(cmd, ["-help"], { timeout: 5_000 });
cachedCairePath = cmd;
return cmd;
} catch {
// try next
}
}
throw new Error(
"caire binary not found. Install via: go install github.com/esimov/caire/cmd/caire@v1.5.0",
);
}
/**
* Content-aware resize using caire (Go seam carving engine).
* Supports both shrinking and enlarging via seam removal/insertion.
*/
export async function seamCarve(
inputBuffer: Buffer,
outputDir: string,
options: SeamCarveOptions = {},
onProgress?: ProgressCallback,
): Promise<SeamCarveResult> {
const inputPath = join(outputDir, "input_seam_carve.png");
const outputPath = join(outputDir, "output_seam_carve.png");
const cairePath = await findCaire();
const id = randomUUID();
const inputPath = join(outputDir, `caire-in-${id}.png`);
const outputPath = join(outputDir, `caire-out-${id}.png`);
await writeFile(inputPath, inputBuffer);
const { stdout } = await runPythonWithProgress(
"seam_carve.py",
[inputPath, outputPath, JSON.stringify(options)],
{ onProgress },
);
try {
await writeFile(inputPath, inputBuffer);
const result = JSON.parse(stdout);
if (!result.success) {
throw new Error(result.error || "Content-aware resize failed");
// Build caire arguments
const args = ["-in", inputPath, "-out", outputPath, "-preview=false"];
if (options.square) {
// Caire -square requires -width and -height set to the shortest edge
const meta = await sharp(inputBuffer).metadata();
const shortest = Math.min(meta.width ?? 0, meta.height ?? 0);
args.push("-square", "-width", String(shortest), "-height", String(shortest));
} else {
if (options.width) args.push("-width", String(options.width));
if (options.height) args.push("-height", String(options.height));
}
if (options.protectFaces) args.push("-face");
if (options.blurRadius !== undefined) args.push("-blur", String(options.blurRadius));
if (options.sobelThreshold !== undefined) args.push("-sobel", String(options.sobelThreshold));
await execFileAsync(cairePath, args, { timeout: 60_000 });
const buffer = await readFile(outputPath);
const meta = await sharp(buffer).metadata();
return {
buffer,
width: meta.width ?? 0,
height: meta.height ?? 0,
};
} finally {
await rm(inputPath, { force: true }).catch(() => {});
await rm(outputPath, { force: true }).catch(() => {});
}
const buffer = await readFile(outputPath);
return {
buffer,
width: result.width,
height: result.height,
};
}
-1
View File
@@ -390,5 +390,4 @@ export const PYTHON_SIDECAR_TOOLS = [
"blur-faces",
"erase-object",
"ocr",
"content-aware-resize",
] as const;