Hindi OCR Image Testing and Reporting

₹600-1500 INR

Posted 1 minute ago

•

Ends in 6 days

₹600-1500 INR

Paid on delivery

Hi, I need the following work done for my project: 1. Collect approximately 1,000 Hindi document images exclusively from the e-Sansad Library website, covering different document categories and image-quality conditions. 2. Test these images using the existing batch-processing feature of my OCR Readiness Platform. No new batch-processing functionality is required. 3. Identify low-scoring factors for each image and apply the existing refinement functionality wherever required. 4. Record the before-and-after scores for all 10 factors, keeping the after-refinement fields blank for factors that were not refined. 5. Prepare an Excel report containing image IDs/names, initial scores, refined scores, factors selected for refinement, and overall scores before and after refinement. 6. Include summary statistics showing how many images had low scores for each factor, how many were successfully refined, and how many showed improvement. The main task is collecting the 1,000 Hindi images, testing them through the existing platform, and preparing the complete Excel report. Please do not modify the existing scoring formulas, factor weights, refinement logic, or other platform features. The batch-processing functionality is already available.

Project ID: 40815145

About the project

1 proposal

Open for bidding

Remote project

Active 14 secs ago

Place your bid

Bid amount Email address

Benefits of bidding on Freelancer

Set your budget and timeframe

It's free to sign up and bid on jobs

1 freelancer is bidding on average ₹930 INR for this job

User Avatar
Flag of INDIA

The main challenge is making the results for roughly 1,000 e-Sansad Library images traceable and comparable—not just running them through OCR readiness checks. The sample needs to cover different document categories and image-quality conditions, while the report must clearly distinguish initial scores, factors selected for refinement, and actual post-refinement results without changing the platform’s existing behavior. I’ll collect the images exclusively from the specified library and keep each image’s ID or filename linked to its source information. I’ll process them using the existing batch feature, review the scores across all 10 factors, and use the existing refinement function only where a factor needs it. I’ll leave after-refinement cells blank for factors that were not refined, as requested, and retain the overall before-and-after scores for each image. The Excel workbook will include image identifiers, initial and refined factor scores, selected refinement factors, and overall scores. I’ll also add summary counts for low scores by factor, successful refinements, and images that improved, keeping refinement success separate from score improvement so the results are easy to interpret. Before delivery, I’ll check that image records align with the platform results and that the summaries match the underlying rows. I’ll leave scoring formulas, weights, refinement logic, and other platform features untouched. How does the platform expose the image IDs and per-factor results—through an export, or should these be recorded from the interface?

内容来自 Freelancer News,版权归原作者所有。

阅读原文 ↗

把喜欢的内容,留在你的新标签页。

打开我的订阅