Incorrect Detection of Rubʻ Marker as Ayah Marker Leads to Inaccurate Ayah Count

[MaZED-UP]

Assalam Alaykum,

I would like to express my gratitude for the effort put into developing this Quran Mushaf AI model. Your dedication to this project is commendable, Jazak Allah Khaira.

After installing the necessary requirements, I ran main.py using these provided Mushaf pages and reviewed the results. I discovered that the model incorrectly identifies a Rubʻ marker as an Ayah marker, causing the Ayah count to be incremented by 1 (right after Ayah #43 in the following sample).

[jrahim]

Walaikumassalam!

You can filter out using the box thresholds. Maybe set a threshold of 0.6 to filter out lower confidence predictions. The rub marker prediction has a confidence of 0.41.

On Wed., May 10, 2023, 3:53 p.m. MaZED-UP, @.***> wrote:

Assalam Alaykum,

I would like to express my gratitude for the effort put into developing this Quran Mushaf AI model. Your dedication to this project is commendable, Jazak Allah Khaira.

After installing the necessary requirements, I ran main.py using these provided Mushaf https://mega.nz/folder/beZxmLgQ#t8yK8lebLPPTL51d_uNPkw pages and reviewed the results. I discovered that the model incorrectly identifies a Rubʻ marker as an Ayah marker, causing the Ayah count to be incremented by 1 (right after Ayah #43 in the following sample).

[image: 6] https://user-images.githubusercontent.com/59477548/237530042-2c33feec-a863-4a97-897e-46abb42cb20c.jpg

— Reply to this email directly, view it on GitHub https://github.com/bllfoad/ayahai/issues/1, or unsubscribe https://github.com/notifications/unsubscribe-auth/AE2NZFCV22R7WT7K2GGZ2D3XFQMADANCNFSM6AAAAAAX5KGFFE . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[MaZED-UP]

I made a modification to main.py on line 11: model.conf = 0.6. However, on page 17, there was an issue with a Rubʻ marker located on the top right. It was mistakenly detected as an Ayah marker and the green annotation was cut off, so I couldn't determine its prediction confidence value.

To address this problem, I increased the value to 0.8, but unfortunately, the issue persisted. Thankfully, setting the value to 0.9 resolved the problem successfully. Alhamdulillah!

Thank you for your assistance throughout this process.