CVF Open Access Paper Stamping Pipeline

Overview

This is cvf-openaccess-publishing-pipeline, a Python CLI tool that processes PDF files for the CVF Open Access Archive

Invocation

# For help:
$ uv run cvf-openaccess-publishing-pipeline --help
$ uv run cvf-openaccess-publishing-pipeline run --help

# Stamp the workshop track:
$ uv run cvf-openaccess-publishing-pipeline run /path/to/cvpr2026-workshops.toml
$ uv run cvf-openaccess-publishing-pipeline run /path/to/cvpr2026-workshops.toml -n # dry run

# First 10 papers:
$ uv run cvf-openaccess-publishing-pipeline run /path/to/cvpr2026-workshops.toml 1-10

# Print as JSON:
$ uv run cvf-openaccess-publishing-pipeline show-json /path/to/cvpr2026-workshops.toml
$ uv run cvf-openaccess-publishing-pipeline show-json /path/to/cvpr2026-workshops.toml | jq .title
$ uv run cvf-openaccess-publishing-pipeline show-json /path/to/cvpr2026-workshops.toml 1-10 | jless # accepts slices

# Generate QA reports:
$ uv run cvf-openaccess-publishing-pipeline report /path/to/cvpr2026-workshops.toml

# Start IPython shell with paper loaded for debugging:
$ uv run cvf-openaccess-publishing-pipeline shell /path/to/cvpr2026-workshops.toml

Inputs for each track

You must prepare:

1. config.toml, a handwritten configuration file
2. A directory full of PDF files from OpenReview which have been prepared by the IEEE Computer Society's Conference Publishing Services (CPS) division
3. A spreadsheet with metadata about the papers from CPS

Outputs for each track

This tool produces:

1. A directory of PDF files
2. papers.csv containing metadata about the papers
3. report.html, a report of thumbnails of every processed paper for manual review

Artifacts 1 and 2 are sent to the CVF Open Access website admin team to be uploaded to the Open Access website. Artifact 3 should be inspected closely by the person running this tool to ensure that the PDF files have been processed correctly.

What this tool does

• Reads the conference information, spreadsheet URL, etc from config.toml
• Reads title, authors, abstract, first/last page, etc. from each row of the spreadsheet
• Verify the following invariants for the input:
  • Every paper listed in the spreadsheet has a corresponding PDF file in the input directory
  • Every PDF file in the input directory has a corresponding row in the spreadsheet
  • Output filenames generated from the spreadsheet metadata are unique and don't collide with each other
• For each paper:
  • Apply PDF metadata: set dc:title to the paper title, dc:creator to the list of paper authors, and dc:description to the long name of the conference
  • Verify page count matches the spreadsheet
  • Stamp the CVF Open Access banner on the first page
  • Stamp page numbers: Apply heuristics to guess whether the author forgot to turn off page numbering in their PDF, and if so, apply a stamp to the bottom of each page with the correct page number from the spreadsheet.
  • Strip author-applied review markup annotations (eg. from Preview.app)
  • Write the processed PDF to the output directory with a suitable output filename
• Generate output spreadsheet, saved to papers.csv
• Generate thumbnail report for manual review, saved to report.html
• Verify the following invariants for the output:
  • papers.csv must exist
  • All output files exist
  • All output files have the same number of pages as on the spreadsheet

What to do next

• Zip up the entire output directory and send it to the CVF Open Access website admin team.

Configuration

Preparing input files

Example directory tree for CVPR 2026, which has "main", "findings", and "workshops" tracks:

Google Drive/My Drive/Areas/CVF Open Access Proceedings/CVPR 2026/
│   
│   Config files for the three tracks:
│   
├── cvpr2026-findings.toml
├── cvpr2026-main.toml
├── cvpr2026-workshops.toml
│   
│   Input papers from IEEE CPS:
│   
├── Files for CVF - CVPR, CVPRW, CVPRF
│   ├── CVPR 2026 - Main Conference
│   │   ├── CVPR 2026 - Metadata for CVF - 260501.gsheet
│   │   ├── CVPR 2026 - Paper PDF Files
│   │   │   ├── 30593.pdf
│   │   │   ├── 30595.pdf
│   │   │   ├── 30599.pdf
│   │   │   │    ⋮
│   │   │   └── 46780.pdf
│   │   └── CVPR 2026 - Supplemental Files
│   │       ├── 30593_supp_1.zip
│   │       ├── 30595_supp_1.pdf
│   │       ├── 30599_supp_1.pdf
│   │       │    ⋮
│   │       └── 46778_supp_1.zip
│   │  
│   ├── CVPRF 2026 - Findings
│   │   ├── CVPRF 2026 - Metadata for CVF - 260508.gsheet
│   │   ├── CVPRF 2026 - Paper PDF Files
│   │   │   ├── 30623.pdf
│   │   │   ├── 30627.pdf
│   │   │   ├── 30629.pdf
│   │   │   │    ⋮
│   │   │   └── 46728.pdf
│   │   └── CVPRF 2026 - Supplemental Files
│   │       ├── 30623_supp_1.zip
│   │       ├── 30627_supp_1.pdf
│   │       ├── 30629_supp_1.pdf
│   │       │    ⋮
│   │       └── 46728_supp_1.zip
│   │  
│   └── CVPRW 2026 - Workshops
│       ├── 5 25 2026 - CVPRW 2026, minus Missing Supps - Metadata for CVF - 260512.gsheet
│       ├── CVPRW 2026 - Metadata for CVF - 260512.gsheet
│       ├── CVPRW 2026 - Metadata for CVF - 260527.gsheet
│       ├── CVPRW 2026 - Paper PDF Files
│       │   ├── 3DMV-10.pdf
│       │   ├── 3DMV-11.pdf
│       │   ├── 3DMV-12.pdf
│       │   │    ⋮
│       │   ├── 6thAdvML@CV-1.pdf
│       │   ├── 6thAdvML@CV-12.pdf
│       │   ├── 6thAdvML@CV-15.pdf
│       │   │    ⋮
│       │   ├── A2A-MML-13.pdf
│       │   ├── A2A-MML-18.pdf
│       │   ├── A2A-MML-19.pdf
│       │   │    ⋮
│       │   ├── A4VM-13.pdf
│       │   ├── A4VM-16.pdf
│       │   ├── A4VM-18.pdf
│       │   │    ⋮
│       │   ├── ABAW-12.pdf
│       │   ├── ABAW-15.pdf
│       │   ├── ABAW-18.pdf
│       │   │    ⋮
│       │   └── XAI4CV-4.pdf
│       └── CVPRW 2026 - Supplemental Files
│           ├── 3DMV-10_supp_1.pdf
│           ├── 3DMV-11_supp_1.pdf
│           ├── 3DMV-16_supp_1.pdf
│           │    ⋮
│           ├── 6thAdvML@CV-1_supp_1.pdf
│           ├── 6thAdvML@CV-2_supp_1.pdf
│           ├── 6thAdvML@CV-4_supp_1.pdf
│           │    ⋮
│           ├── A2A-MML-13_supp_1.pdf
│           ├── A2A-MML-20_supp_1.pdf
│           │    ⋮
│           └── XAI4CV-4_supp_1.pdf
│   
│   Outputs written by this tool:
│   
├── output-findings-20260516
│   ├── papers.csv
│   ├── Abdelgawad_Online_Interpretable_Matrix_Decomposition_for_Large-Scale_Streaming_Data_CVPRF_2026_paper.pdf
│   ├── Abid_Gazemo_Mimicking_Human_Saccades_via_Foveal-Peripheral_Feature_Modeling_for_Lightweight_CVPRF_2026_paper.pdf
│   ├── Aboukhadra_GHOST_Fast_Category-Agnostic_Hand-Object_Interaction_Reconstruction_from_RGB_Videos_Using_CVPRF_2026_paper.pdf
│   │    ⋮
│   └── Zuo_Channel_Correlation_Loss_for_Binary_Neural_Networks_CVPRF_2026_paper.pdf
├── output-main-20260516
│   ├── papers.csv
│   ├── Abdal_Visual_Personalization_Turing_Test_CVPR_2026_paper.pdf
│   ├── Abdelfattah_OSMO_Open-vocabulary_Self-eMOtion_Tracking_CVPR_2026_paper.pdf
│   ├── Abousamra_TopoSlide_Topologically-Informed_Histopathology_Whole_Slide_Image_Representation_Learning_CVPR_2026_paper.pdf
│   │    ⋮
│   └── Zuo_SketchRevive_Fine-Grained_Pixel-to-Vector_Sketch_Completion_with_Diffusion-Prior-Guided_Multimodal_LLMs_CVPR_2026_paper.pdf
├── output-workshops-20260516
│   ├── papers.csv
│   ├── w1
│   │   ├── Gogawale_Bag_of_Bags_Adaptive_Visual_Vocabularies_for_Genizah_Join_Image_CVPRW_2026_paper.pdf
│   │   └── Yin_Documentation_Infrastructure_and_Ethical_Challenges_in_Spatial_AI_for_Cultural_CVPRW_2026_paper.pdf
│   ├── w10
│   │   ├── Bhatia_Li-AutoFlow_Autoregressive_Flow_Matching_for_Continuous_AV_Scene_Prediction_CVPRW_2026_paper.pdf
│   │   ├── Chahe_Policy-Guided_World_Model_Planning_for_Language-Conditioned_Visual_Navigation_CVPRW_2026_paper.pdf
│   │   ├── Chen_Intelligent_Robot_Manipulation_Requires_Self-Directed_Learning_CVPRW_2026_paper.pdf
│   │   │    ⋮
│   │   └── Yaman_Remedying_the_Curse_of_Autonomous_Driving_VLM_Driven_Training-Free_Framework_CVPRW_2026_paper.pdf
│   ├── w11
│   │   ├── Chiu_Edge-Efficient_Vision-Language_Models_for_Autonomous_Driving_Using_Distillation_and_RAG-Based_CVPRW_2026_paper.pdf
│   │   ├── Haklidir_When_Does_Adaptive_Guidance_Help_Belief-Aware_Privileged_Distillation_for_Autonomous_CVPRW_2026_paper.pdf
│   │   ├── Lengyel_CCLSTM_Coupled_Convolutional_Long-Short_Term_Memory_Network_for_Occupancy_Flow_CVPRW_2026_paper.pdf
│   │   │    ⋮
│   │   └── Yong_Localization-Guided_Foreground_Augmentation_in_Autonomous_Driving_CVPRW_2026_paper.pdf
│   ├── w12
│   │   ├── Holzemann_SCAR_Satellite_Imagery-Based_Calibration_for_Aerial_Recordings_CVPRW_2026_paper.pdf
│   │   ├── Jozsa_RF-Loc_Robust_Visual-Radio_Frequency_Localization_via_Hierarchical_Point_Cloud_Registration_CVPRW_2026_paper.pdf
│   │   ├── Mandal_VGGT-SLAM_CVPRW_2026_paper.pdf
│   │   └── Safavigerdini_Gram-Schmidt_Feature_Reduction_for_Local_Feature_Descriptor_Compression_CVPRW_2026_paper.pdf
│   │    ⋮
│   │    ⋮
│   │    ⋮
│   └── w97
│       ├── Gui_Object-Aware_4D_Human_Motion_Generation_CVPRW_2026_paper.pdf
│       ├── hari_SimScene_Automated_Photorealistic_Scene_Reconstruction_for_Geographically_Scalable_Physical_AI_CVPRW_2026_paper.pdf
│       ├── Karhade_Any4D_Unified_Feed-Forward_Metric_4D_Reconstruction_CVPRW_2026_paper.pdf
│       │    ⋮
│       └── Zhang_3D_Gaussian_Splatting_for_Efficient_Retrospective_Dynamic_Scene_Novel_View_CVPRW_2026_paper.pdf
└── qa-reports
    ├── findings-20260516-a-00001-of-00001.html
    ├── findings-20260516-a-00001-of-00004.html
    ├── findings-20260516-a-00002-of-00004.html
    ├── findings-20260516-a-00003-of-00004.html
    ├── findings-20260516-a-00004-of-00004.html
    ├── main-20260516-a-00001-of-00014.html
    ├── main-20260516-a-00002-of-00014.html
    ├── main-20260516-a-00003-of-00014.html
    ├── main-20260516-a-00004-of-00014.html
    │    ⋮
    ├── main-20260516-a-00014-of-00014.html
    ├── workshops-20260516-a-00001-of-00001.html
    ├── workshops-20260516-a-00001-of-00004.html
    ├── workshops-20260516-a-00002-of-00004.html
    │    ⋮
    └── workshops-20260527-a-00004-of-00004.html

Example configuration file for CVPR 2026 main track (cvpr2026-main.toml):

acronym            = "CVPR"
long_name          = "IEEE Conference on Computer Vision and Pattern Recognition"
year               = 2026

input_path         = "Files for CVF - CVPR, CVPRW, CVPRF/CVPR 2026 - Main Conference/CVPR 2026 - Paper PDF Files/"
output_path        = "output-main-20260516/"
report_output_path = "qa-reports/main-20260516-a.html"

[banner]
text = """
This CVPR paper is the Open Access version, provided by the Computer Vision Foundation.
Except for this watermark, it is identical to the accepted version;
the final published version of the proceedings is available on IEEE Xplore.
"""

[spreadsheet]
url = "https://docs.google.com/spreadsheets/d/.../export?format=csv"
# The above Google Sheets doc is shared to the public ("Anyone with the link can access").
# We manually edited the URL to add `/export?format=csv` at the end.
# This URL links to a CSV export of the spreadsheet, which is what
# this tool expects as input.

# The list of column names (Row 1 in the google spreadsheet)
# Sometimes these change from year to year
[spreadsheet.columns]
title                 = "Title (Corrected)"
abstract              = "Abstract (Cleaned)"
authors               = "Authors (Corrected)"
authors_separator     = ","
input_filename        = "Camera-Ready File (Corrected)"
supplemental_filename = "Supplemental File (Corrected)"
first_page            = "First Page"
last_page             = "Last Page"

Example configuration file for CVPR 2026 workshop track (cvpr2026-workshops.toml)

acronym            = "CVPRW"
long_name          = "IEEE Conference on Computer Vision and Pattern Recognition Workshops"
year               = 2026
input_path         = "Files for CVF - CVPR, CVPRW, CVPRF/CVPRW 2026 - Workshops/CVPRW 2026 - Paper PDF Files/"
output_path        = "output-workshops-20260527/"
zip_output_path    = "output-workshops-20260527.zip"
report_output_path = "qa-reports/workshops-20260527-a.html"

[banner]
text = """
This CVPR Workshop paper is the Open Access version, provided by the Computer Vision
Foundation. Except for this watermark, it is identical to the accepted version;
the final published version of the proceedings is available on IEEE Xplore.
"""

[spreadsheet]
url = "https://docs.google.com/spreadsheets/d/.../export?format=csv"

[spreadsheet.columns]
title                  = "Title (Corrected)"
abstract               = "Abstract (Cleaned)"
authors                = "Authors (Corrected)"
authors_separator      = ","
input_filename         = "Camera-Ready File (Corrected)"
supplemental_filename  = "Supplemental File (Corrected)"
first_page             = "First Page"
last_page              = "Last Page"
output_filename_prefix = "Output Filename Prefix"

Note the followng differences between the main and workshop track:

• The acronym and long_name fields now mention workshops
• The input_path field has been updated to point to the correct directory of PDF files
• All three output_paths have been changed
• The banner text refers to "CVPR Workshop paper" instead of just "CVPR paper", and has been wrapped differently
• The [spreadsheet].url points to the workshop spreadsheet
• There's an output_filename_prefix field in the [spreadsheet.columns] section. The spreadsheet contains cells like w1/, w3/, w5/, etc., which are used as prefixes for the output filenames to ensure that they are unique and to group papers by workshop in the output directory. The tool will insert this prefix at the beginning of each output filename, before the paper title and conference acronym. For example, if the original output filename would have been Gogawale_Bag_of_Bags_Adaptive_Visual_Vocabularies_for_Genizah_Join_Image_CVPRW_2026_paper.pdf, and the output_filename_prefix value for that row is w1/, then the final output filename will be w1/Gogawale_Bag_of_Bags_Adaptive_Visual_Vocabularies_for_Genizah_Join_Image_CVPRW_2026_paper.pdf.

CVF / IEEE organizational structure

This tool manages all of the blue parts of the following flowchart: