List Formats Command

​

Usage

panlabel list-formats

​

Output

• FORMAT - The canonical format name
• READ - Whether Panlabel can read this format
• WRITE - Whether Panlabel can write this format
• LOSSINESS - How much information is preserved relative to Panlabel’s IR
• DESCRIPTION - Brief description of the format

​

Example Output

Supported formats:

  FORMAT       READ   WRITE  LOSSINESS    DESCRIPTION
  ------       ----   -----  ---------    -----------
  ir-json      yes    yes    lossless     Panlabel's intermediate representation (JSON)
  coco         yes    yes    lossless     COCO object detection format (JSON)
  cvat         yes    yes    conditional  CVAT for images XML annotation export
  label-studio yes    yes    conditional  Label Studio task export (JSON)
  tfod         yes    yes    lossy        TensorFlow Object Detection format (CSV)
  yolo         yes    yes    lossless     Ultralytics YOLO .txt (directory-based)
  voc          yes    yes    lossless     Pascal VOC XML (directory-based)

Lossiness key:
  lossless    - Format preserves all IR information
  conditional - Format may lose info depending on dataset content
  lossy       - Format always loses some IR information

Tip: Use '--from auto' with 'convert' for automatic format detection.

​

Lossiness Categories

​

Lossless

• ir-json - Panlabel’s native format (by definition lossless)
• coco - COCO JSON preserves images, annotations, categories, and metadata
• yolo - YOLO with data.yaml preserves all required information
• voc - Pascal VOC XML preserves bounding boxes and metadata

​

Conditional

• cvat - Conditionally lossy based on:
  • Presence of annotation attributes not supported by CVAT
  • Dataset-level metadata that CVAT doesn’t represent
  • Images without annotations (may be handled differently)
• label-studio - Conditionally lossy based on:
  • Task-level metadata structure
  • Annotation attributes specific to Label Studio’s schema
  • Image hosting and reference method

​

Lossy

• tfod (TensorFlow Object Detection CSV) - Always lossy because:
  • No support for dataset-level metadata
  • Images without annotations are dropped
  • Limited category metadata (only ID and name)
  • No support for annotation attributes
  • No bounding box attributes beyond coordinates

​

Format Details

​

ir-json

• Canonical internal format
• Preserves all information
• Human-readable JSON structure
• Use as a universal interchange format

​

coco

• Industry-standard format
• Single JSON file with all annotations
• Widely supported by ML frameworks
• Aliases: coco-json

​

cvat

• XML-based format from CVAT annotation tool
• Can be single file or directory export
• Supports rich annotation attributes
• Aliases: cvat-xml

​

label-studio

• JSON format from Label Studio annotation platform
• Array of task objects
• Includes annotation metadata and provenance
• Aliases: label-studio-json, ls

​

tfod

• Simple CSV format
• One row per annotation
• Limited metadata support
• Aliases: tfod-csv

​

yolo

• Directory structure with labels/ subdirectory
• One .txt file per image with normalized coordinates
• Requires classes.txt and data.yaml for category info
• Aliases: ultralytics, yolov8, yolov5

​

voc

• XML format from Pascal VOC challenge
• Directory structure with Annotations/ and JPEGImages/
• One XML file per image
• Aliases: pascal-voc, voc-xml

​

Using with Other Commands

​

Convert Command

# Use format names from list-formats
panlabel convert --from coco --to yolo -i input.json -o ./output/

​

Validate Command

panlabel validate dataset.xml --format cvat

​

Stats Command

panlabel stats dataset/ --format yolo

​

Sample Command

panlabel sample -i input.json -o output.json --from label-studio --to coco

​

Checking Format Compatibility

1. Which formats support your use case: All listed formats support both read and write
2. Potential information loss: Check the lossiness column before converting
3. Available aliases: Use shorter names like ls instead of label-studio-json

​

Planning Conversions

# Check lossiness before converting
panlabel list-formats | grep tfod
# Output: tfod ... lossy ...

# Convert with awareness of potential loss
panlabel convert -f coco -t tfod -i input.json -o output.csv --allow-lossy

​

Integration Examples

​

Scripted Format Discovery

#!/bin/bash
# Get all supported formats programmatically
panlabel list-formats | tail -n +4 | head -n -5 | awk '{print $1}'

​

CI/CD Format Validation

# Verify a format is supported before processing
FORMAT="yolo"
if panlabel list-formats | grep -q "^  $FORMAT "; then
  echo "Format $FORMAT is supported"
else
  echo "Format $FORMAT is not supported"
  exit 1
fi

​

Tips

​

See Also