Target List : https://docs.google.com/document/d/1TFQ8-SHsg84bS4CRZpPJWmkuNW2altRtZtIgUO5Cfjk/edit

Bhaashini API Targets

Input : JSON file consisting scribbles from Stage 1 Output : JSON File consisted of detected polygons , scribbles

Note : It should only contain the corrections from the existing pipeline and not any additional changes in the design. | @Anonymous | P0 | @ **Re-Scribble Generation : NV - Not curved + not missing**
| | | | | [ 1 ] Bare Flask Application [V0]

Mode : single/multiple images

For starters we can have 3 starting points :

0.line-segmentation-api/visualise_single_image : JSON File + Visualise Polygons + Visualise Scribbles [MODE : SINGLE only]

1. line-segmentation-api/generate_polygons : JSON File to Download
2. line-segmentation-api/visualise : Two Options - Scribbles + Polygons

In all the above apis , they will internally call bash files or something more optimal to get the results.

Note : Here are the overall drop-downs :

1. Mode - single/multiple < Not Required , can be detected backend >
2. Model - I2/BL/KH/SD
3. Visualise - Scribbles/Polygons/Both/None
4. Export - JSON [ Default ]

• Complete Cmdlet mode should be supported and tested . | @Anonymous | P1 | @Niharika should provide the entire Stage1 setup with outputs written to JSON file.

[ But the entire setup can be up with assuming I2_TEST.json as initial input] | | | | | [ 2 ] Flask Application : Additional

1. Any input image / input folder of images must be stored in the backed also with the outputs provided to them .

• Helps us to debug the performance on new images.
• Helps us to understand the use cases in efficient manner.

To store :

1. Scribbles 2.Binarised Image 3.Seams (upper + lower) 4.Polygons

5.Time of Request Recieved , Total Time Tracking Store ⇒ zip file / pickle (store efficiently ) | @Anonymous | P2 | None | | | | | [ 3 ] CI / CD Pipeline

• Update the hosted API code with the last stable version of the SeamFormer Codebase. | @Anonymous + @Anonymous | <TBD> | <TBD> | | | |

Stage 1 Targets : Public Repo

Dataset Streamline

1. Removal of images with no text lines in them. 2. Erasing out the figures in the final set of images. [ Discussion ]

KHMER Dataset : 1. Can we generate the binary images for remaining data of text line ?

I2 Dataset :

1. Remove / Add the images we are currently not able to access in code and throws exceptions. | @Waseem | P0 | None | InProgress : Need final feedback from Sir . | 29 July EOD. | | [1] Re-Scribble Generation :

2. For the given list of indices , certain scribbles are missing , either remove these images or try to generate . ( -1 scribble is acceptable) [ For I2 + BKS , perform this testing ]

Note : All such cases must be listed in a text file and maintained for future reference . ImageFile-ParentDataset.

| @Niharika + Waseem | P0 | None

< Will skip removing all the scribble generation for the removed Bhoomi samples > | InProgress

• Completed for I2_TEST. | 29 July EOD. | | [2] Google Drive : Upload

Uploading all the required datasets - binary and scribbles , and gd polygons to Google Drive for one stop download. | @Niharika | P0 | [1] & [2] from above . | | 29 July EOD. |

Binarisation Branch

Based on the new training splits , and corrections to binary gnd truths , we might need to take a call on retraining. | @Waseem | P1 | Above dataset inspection. | Done | 29 July 2023 |

Scribble Branch

Stage II Targets : Public Repo

Need visual + metric analysis of the new Colab version of Stage II notebook. | @Anonymous | P0 | None ( Vaibhav has already authored the notebook) | | 30th July EOD | | [1] Intermediate Folder of the results

The user need or need not choose to visualise the output images | @Anonymous | P0 | @Sreenya has to perform evaluation. | | | | [2] Documentation Work | Primary : @Anonymous Secondary : Vaibhav Agrawal | P1 | Based on Vaibhav’s final version of the notebook. | Started | 31st July EOD |

Teaser Work

Need to add content for each of these slides**

Slides : [V0]

• Input / Output
• Overall Pipeline
• Stage 1 Diagram
• Stage 2 Diagram
• Comparison of Palmira and SeamFormer (IndiscapesV1 ? )
• Result Table

( < = 1 min ) | Primary : Niharika Secondary: < Sreenya / Arnav > | P0 | None

• Require the Palmira Template of the presentation. | NotStarted | | | Animations

Animations : GIF - Upper & Lower Seam Generation for one line

**** | | P0 | None | NotStarted | | | Automated Voice for Teaser Video

Discussion on the tool that was used last time for Palmira , any further preferences ? **** | TBD. < Sreenya > | P0 | None | Not Started | | | Arxiv Paper Edits

• What should be the corrections from the rebuttal paper ? [ You can add more details , no restrictions , rebuttal paper will be baseline ] - more qual images . | Niharika | P0 | Discussion with Sir | NotStarted | TBD. |

Poster Work