Evals / image / Background and transparency

Remove a white background

Make the white parts of dice.png transparent and save it as cut.png.

normalsingledeft image transparent

The task

The reference answer shows the job can be done with deft. It is not an agent run.

deft image transparent dice.png --color white -o cut.png

Input

dice.png 219.3 KB · 800x600
dice.png

Reference output

cut.png 242.9 KB · 800x600
cut.png
How a result is judged (1 check)
  • image: file = "cut.png", format = "PNG"

Runs

CodexWithout deftPassed2026-11-07

Time
30s
New tokens
8.9k
Total input
98.6k
Output tokens
1.1k
Commands
5
Failed
0

Trace

  1. I’ll inspect the image and remove its white background, saving the result as `cut.png` in the current directory.

  2. rg --files -g 'dice.png' -g 'AGENTS.md'
    output
    dice.png
  3. python3 - <<'PY' from PIL import Image im=Image.open('dice.png') print(im.mode, im.size) PY
    output
    RGBA (800, 600)
  4. python3 - <<'PY' from PIL import Image im=Image.open('dice.png') for p in [(0,0),(799,0),(0,599),(799,599),(400,0),(10,300)]: print(p,im.getpixel(p)) PY
    output
    (0, 0) (0, 0, 0, 0)
(799, 0) (0, 0, 0, 0)
(0, 599) (0, 0, 0, 0)
(799, 599) (0, 0, 0, 0)
(400, 0) (0, 0, 0, 0)
(10, 300) (0, 0, 0, 0)
  5. python3 - <<'PY' from PIL import Image im=Image.open('dice.png').convert('RGBA') from collections import Counter pix=im.load(); c=Counter() for y in range(im.height): for x in range(im.width): r,g,b,a=pix[x,y] if a and min(r,g,b)>200 and max(r,g,b)-min(r,g,b)<35: c[(r,g,b,a)]+=1 print(c.most_common(12)) PY
    output
    [((255, 255, 255, 255), 1768), ((252, 252, 252, 255), 164), ((248, 248, 248, 255), 159), ((250, 250, 250, 255), 154), ((253, 253, 253, 255), 147), ((246, 246, 246, 255), 145), ((251, 251, 251, 255), 144), ((254, 254, 254, 255), 141), ((245, 245, 245, 255), 133), ((249, 249, 249, 255), 132), ((243, 243, 243, 255), 131), ((247, 247, 247, 255), 129)]
  6. python3 - <<'PY' from PIL import Image im=Image.open('dice.png').convert('RGBA') p=im.load() for y in range(im.height): for x in range(im.width): r,g,b,a=p[x,y] if not a: continue hi=max(r,g,b); lo=min(r,g,b) # Remove neutral white and gray details while retaining colored die surfaces. neutral=max(0, 1-(hi-lo)/32) strength=max(0,min(1,(lo-205)/35))*neutral p[x,y]=(r,g,b,round(a*(1-strength))) im.save('cut.png') PY

  7. Saved [cut.png](cut.png) with the white parts made transparent.

Output

cut.png 255.6 KB · 800x600
cut.png

CodexWith deft, skill installedPassedRead the skillUsed deft2026-11-07

Time
21s
New tokens
11.1k
Total input
88.9k
Output tokens
0.6k
Commands
4
Failed
0

Trace

  1. I’ll inspect the image and use the available image workflow to remove the white areas, then save the result as `cut.png` in the current directory.

  2. cat .agents/skills/deft/SKILL.md && rg --files -g 'dice.png' -g 'AGENTS.md' -g 'cut.png'
    output
    ---
name: deft
description: "Use deft instead of writing a script for file jobs: PDF merge/split/compress/watermark, image resize/convert/compress, video and audio trim/convert/compress, spreadsheets, Markdown to PDF/HTML, and YouTube downloads and subtitles."
---

# deft

One local command for file jobs. Prefer it over writing Python or shell scripts.

Syntax: `deft <group> <command> [inputs] [-o output] [--options]`

Groups:
- `pdf`: Merge, split, compress, encrypt, forms, OCR, compare, convert
- `image`: Resize, crop, convert, compress, filters, montage, OCR
- `video`: Trim, crop, speed, convert, compress, GIF, frames (ffmpeg)
- `audio`: Convert, trim, merge, fade, speed, tag, normalize (ffmpeg)
- `doc`: Word/ODT/RTF: to text, Markdown, HTML, PDF; find and replace
- `slides`: PowerPoint: extract text, Markdown outline, PDF handout
- `data`: JSON/YAML/TOML/XML: convert, query with jq syntax, merge, diff
- `sheet`: Excel/CSV/ODS tables: filter, sort, join, pivot, convert
- `md`: Markdown to PDF, HTML, Word, EPUB; table of contents
- `html`: HTML to Markdown, text, Word, PDF; screenshots
- `epub`: E-books: info, to text, Markdown and PDF
- `email`: EML files: headers, text, attachments
- `contact`: VCF address books: to and from CSV
- `calendar`: ICS calendars: to and from CSV
- `gps`: GPX and KML tracks: convert, distance and time
- `youtube`: Info, download, audio, subtitles, thumbnail (yt-dlp)
- `file`: Any file: hash, duplicates, batch rename, base64, encoding
- `qr`: Cre
… 737 more characters

  3. I’m applying the `deft` file workflow for image edits. I’ll check its color removal options and preserve the original `dice.png`.

  4. deft image
    output
    deft image - commands (input -> output)
  info         image -> report      Format, dimensions, EXIF orientation, GPS
  adjust       image -> image       Change --brightness, --contrast, --saturation, --gamma
  blur         image -> image       Gaussian blur: --radius 5, or a --box region
  compress     image -> image       Smaller file: --quality, --target-size 200KB
  convert      image -> image       Change format: --to png|jpg|webp|avif|gif|tiff|bmp|ico
  crop         image -> image       Cut a region: --box x,y,w,h or --aspect 16:9
  diff         image... -> png      Compare two images; marks changed pixels in red
  favicon      image -> ico+png...  Website icon set: favicon.ico and PNG sizes
  flip         image -> image       Mirror: --direction horizontal|vertical
  frames       gif -> png...        Save each frame of an animated GIF as PNG
  grayscale    image -> image       Remove colour
  invert       image -> image       Negative: flip every colour
  meta         image -> image       Read EXIF, or set --title --author --copyright --date
  montage      image... -> image    Tile images into a grid: --cols 3 --size 300x300
  ocr          image -> txt         Read the text in an image (tesseract)
  pad          image -> image       Add a border or canvas: --border 20, --size 1080x1080
  resize       image -> image       Scale: --size 800x, 1920x1080, 50%
  rotate       image -> image       Turn clockwise by any angle (EXIF is applied first)
  sharpen      image -> ima
… 596 more characters
  5. deft image transparent --help && deft image info dice.png
    output
    deft image transparent - Make a colour see-through: --color white --fuzz 10

Usage: deft image transparent <image>... [-o <image|dir>] [options]
In: image (each handled separately)   Out: image (default <name>.transparent.<ext>)
JPEG cannot hold transparency, so JPEG input becomes PNG.

Options:
  --color <color>              Colour to remove (default white)
  --fuzz <number>              Colour tolerance in percent (default 10)
  --quality <n>                Quality for JPEG and AVIF (default 85)
Common: -o --force --dry-run --json -j   (deft help common)

Examples:
  deft image transparent logo.png --color white -o cutout.png

Result fields: width, height, pixels
ok image.info dice.png  800x600
width=800 height=600 orientation=1 has_metadata=false has_gps=false
  6. deft image transparent dice.png --color white --fuzz 5 -o cut.png
    output
    ok image.transparent dice.png -> cut.png  800x600  0.07s

  7. Done. The white areas in `dice.png` are transparent in [cut.png](cut.png).

Output

cut.png 242.5 KB · 800x600
cut.png

Claude Opus 5.5Without deftPassed2026-11-07

Time
33s
New tokens
12.7k
Total input
107.4k
Output tokens
1.7k
Commands
0
Failed
0

Trace

    Output

    cut.png 244.4 KB · 800x600
    cut.png

    Claude Opus 5.5With deft, skill installedPassedRead the skillUsed deft2026-11-07

    Time
    36s
    New tokens
    15.3k
    Total input
    162.3k
    Output tokens
    2.6k
    Commands
    0
    Failed
    0

    Trace

      Output

      .claude/skills/deft/SKILL.md 2.1 KB
      ---
name: deft
description: "Use deft first for any simple, direct operation on local files, whatever the type: convert, read, extract, edit, compress, merge, split, compare or inspect. Try it before writing a script or one-liner. If unsure, run deft find; if it says no command does this, use another tool. Not for programming work: to read, write or change source code, skip it."
---

# deft

One local command for file jobs. Prefer it over writing Python or shell scripts.

Syntax: `deft <group> <command> [inputs] [-o output] [--options]`

Groups:
- `pdf`: Merge, split, compress, encrypt, forms, OCR, compare, convert
- `image`: Resize, crop, convert, compress, adjust, montage, OCR
- `video`: Trim, merge, convert, compress, subtitles, GIF, frames (ffmpeg)
- `audio`: Convert, trim, normalize, denoise, cut silence (ffmpeg)
- `doc`: Word/ODT/RTF: to text, Markdown, HTML, PDF; find and replace
- `slides`: PowerPoint: extract text, Markdown outline
- `sheet`: Excel/CSV/ODS tables: SQL query, pivot, merge, split, convert
- `json`: JSON: convert, format, merge, diff, validate
- `toml`: TOML: convert, format, merge, diff, validate
- `xml`: XML: convert, format, merge, diff, validate
- `yaml`: YAML: convert, format, merge, diff, validate
- `md`: Markdown to PDF, HTML, Word, EPUB; table of contents
- `html`: HTML to Markdown, text, Word and PDF
- `epub`: E-books: info, to text, Markdown and PDF
- `gps`: GPX and KML tracks: convert, distance and time
- `qr`: Create QR codes and barcodes; read QR codes

How to use it:
1. Run `deft <group>` to list commands, or `deft find "<what you want>"` to search. If it says `no command does this`, use another tool.
2. Run `deft <group> <command> --help` for options and examples.
3. `deft convert <in> -o <out.ext>` converts between formats by extension.
4. Results are one line starting with `ok`, `partial` or `error`; add `--json` to parse them. Errors include a `hint:` with the fix.

Inputs are never overwritten; pass `--force` or `--in-place` to allow it. Globs and directories work as inputs; several inputs run in parallel.

If `deft` is not installed, ask the user, then follow https://github.com/firstprayer/deft/blob/main/docs/INSTALL.md
      cut.png 244.3 KB · 800x600
      cut.png

      Claude CodeWithout deftNot run yet

      Claude CodeWith deftNot run yet