vsegments 0.1.5 → 0.1.8

package/.claude/settings.local.json

@@ -0,0 +1,13 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(npm rebuild:*)",
+      "Bash(vsegments:*)",
+      "Bash(node -e:*)",
+      "Bash(git add:*)",
+      "Bash(git commit:*)",
+      "Bash(npm publish:*)",
+      "Skill(excalidraw)"
+    ]
+  }
+}

package/bin/cli.js

@@ -11,11 +11,13 @@ const VSegments = require('../src/index');
 
 const program = new Command();
 
+const pkg = require('../package.json');
+
 program
   .name('vsegments')
   .description('Visual segmentation and bounding box detection using Google Gemini AI')
-  .version('0.1.0')
-  .requiredOption('-f, --file <image>', 'Path to input image file')
+  .version(pkg.version)
+  .option('-f, --file <image>', 'Path to input image file')
   .option('--segment', 'Perform segmentation instead of bounding box detection')
   .option('--api-key <key>', 'Google API key (default: GOOGLE_API_KEY env var)')
   .option('-m, --model <model>', 'Model name to use', 'gemini-flash-latest')
@@ -38,6 +40,42 @@ program.parse(process.argv);
 
 const options = program.opts();
 
+// Show welcome message if no file provided
+if (!options.file) {
+  console.log(`
+vsegments v${pkg.version}
+Visual segmentation and bounding box detection using Google Gemini AI
+
+QUICK START
+  vsegments -f image.jpg                   Detect objects with bounding boxes
+  vsegments -f image.jpg --segment         Perform segmentation with masks
+  vsegments -f image.jpg -o output.png     Save visualization to file
+
+CUSTOM PROMPTS
+  vsegments -f photo.jpg -p "find all faces"
+  vsegments -f room.jpg -p "furniture items"
+
+OUTPUT OPTIONS
+  --json results.json    Export detection data as JSON
+  --compact              Print minimal output: "1. label [x y xx yy]"
+  --raw                  Show raw API response
+
+CONFIGURATION
+  --api-key <key>        Google API key (or set GOOGLE_API_KEY env var)
+  --model <name>         Model to use (default: gemini-flash-latest)
+  --temperature <0-1>    Sampling temperature (default: 0.5)
+  --max-objects <n>      Max objects to detect (default: 25)
+
+VISUALIZATION
+  --line-width <n>       Bounding box line width (default: 4)
+  --font-size <n>        Label font size (default: 14)
+  --alpha <0-1>          Mask transparency (default: 0.7)
+
+Run 'vsegments --help' for full options.
+`);
+  process.exit(0);
+}
+
 async function main() {
   try {
     // Validate file exists

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vsegments",
-  "version": "0.1.5",
+  "version": "0.1.8",
   "description": "Visual segmentation and bounding box detection using Google Gemini AI",
   "main": "src/index.js",
   "types": "src/index.d.ts",

package/src/core.js

@@ -41,12 +41,11 @@ class VSegments {
 
     // Default system instructions
     this.defaultSystemInstructions = `
-Return bounding boxes as a JSON array with labels. 
-Never return masks or code fencing. Limit to ${this.maxObjects} objects.
+Return bounding boxes as a JSON array with labels.
+Never return masks or code fencing. Limit to ${this.maxObjects} objects maximum.
 
 Follow this intuition:
-If an object is present multiple times, name them according to their unique characteristic 
-(colors, size, position, unique characteristics, etc..).
+
 If the object is a face number them according to how someone would draw the features of the face. First the left eye, then the right eye
 then the nose, then the left ear, then the right ear, then the mouth, then the chin.
 Same with Animal faces.
@@ -54,6 +53,8 @@ Same with Animal faces.
 General guideline:
 Follow a drawing order intuitively. People usually do not draw first an eye, and then the background and then the shirt but they
 follow a more natural order with symmetry in mind.
+
+Return no more than ${this.maxObjects} features!
     `.trim();
 
     // Safety settings
@@ -77,7 +78,33 @@ follow a more natural order with symmetry in mind.
     // Convert SVG to PNG for API compatibility
     if (mimeType === 'image/svg+xml') {
       const sharp = require('sharp');
-      imageBuffer = await sharp(imageBuffer)
+
+      // Remove common registration/cut line colors from SVG before conversion
+      // These colors are often used for print registration marks, cut lines, etc.
+      let svgString = imageBuffer.toString('utf-8');
+      const registrationColors = [
+        '#ec008c', '#ED008C', // Magenta/pink registration
+        '#00ff00', '#00FF00', // Green registration
+        '#ff0000', '#FF0000', // Red registration (when used for cut lines)
+      ];
+
+      // Remove elements with registration colors
+      for (const color of registrationColors) {
+        // Remove stroke colors
+        const strokeRegex = new RegExp(`stroke="${color}"`, 'gi');
+        svgString = svgString.replace(strokeRegex, 'stroke="none"');
+        // Remove fill colors
+        const fillRegex = new RegExp(`fill="${color}"`, 'gi');
+        svgString = svgString.replace(fillRegex, 'fill="none"');
+      }
+
+      // Increase thin stroke widths for better visibility
+      svgString = svgString.replace(/stroke-width:\s*0\.5pt/gi, 'stroke-width: 2pt');
+      svgString = svgString.replace(/stroke-width="0\.5pt"/gi, 'stroke-width="2pt"');
+
+      imageBuffer = await sharp(Buffer.from(svgString), { density: 300 })
+        .resize(1024, 1024, { fit: 'inside', withoutEnlargement: false })
+        .flatten({ background: { r: 245, g: 245, b: 245 } })  // Light gray background for better contrast
         .png()
         .toBuffer();
       mimeType = 'image/png';
@@ -132,6 +159,7 @@ follow a more natural order with symmetry in mind.
    * @param {Object} options - Detection options
    * @param {string} options.prompt - Custom prompt
    * @param {string} options.customInstructions - Additional system instructions
+   * @param {number} options.temperature - Sampling temperature (overrides constructor value)
    * @param {number} options.maxSize - Maximum image dimension
    * @returns {Promise<SegmentationResult>} - Detection results
    */
@@ -139,6 +167,7 @@ follow a more natural order with symmetry in mind.
     const {
       prompt = 'Detect the 2d bounding boxes',
       customInstructions = null,
+      temperature = this.temperature,
       maxSize = 1024
     } = options;
 
@@ -158,7 +187,7 @@ follow a more natural order with symmetry in mind.
       result = await model.generateContent({
         contents: [{ role: 'user', parts: [{ text: prompt }, image] }],
         generationConfig: {
-          temperature: this.temperature,
+          temperature: temperature,
         }
       });
 
@@ -184,12 +213,14 @@ follow a more natural order with symmetry in mind.
    * @param {string} imagePath - Path to image file
    * @param {Object} options - Segmentation options
    * @param {string} options.prompt - Custom prompt
+   * @param {number} options.temperature - Sampling temperature (overrides constructor value)
    * @param {number} options.maxSize - Maximum image dimension
    * @returns {Promise<SegmentationResult>} - Segmentation results
    */
   async segment(imagePath, options = {}) {
     const {
       prompt = 'Give the segmentation masks for the objects. Output a JSON list of segmentation masks where each entry contains the 2D bounding box in the key "box_2d", the segmentation mask in key "mask", and the text label in the key "label". Use descriptive labels.',
+      temperature = this.temperature,
       maxSize = 1024
     } = options;
 
@@ -213,7 +244,7 @@ follow a more natural order with symmetry in mind.
       result = await model.generateContent({
         contents: [{ role: 'user', parts: [{ text: prompt }, image] }],
         generationConfig: {
-          temperature: this.temperature,
+          temperature: temperature,
         }
       });
 

package/src/index.d.ts

@@ -56,11 +56,13 @@ export interface VSegmentsOptions {
 export interface DetectBoxesOptions {
   prompt?: string;
   customInstructions?: string;
+  temperature?: number;
   maxSize?: number;
 }
 
 export interface SegmentOptions {
   prompt?: string;
+  temperature?: number;
   maxSize?: number;
 }