OCR system

Hello, Dmytro!

The task of automating the recognition of postal envelopes is very interesting and fully implementable in Python using computer vision and OCR tools.

How I will implement this project:

Recognition of indices and text: For printed text, I will set up work with EasyOCR / Tesseract, and for handwritten digits (indices in special boxes or on the margins), I will implement character detection and segmentation for maximum recognition accuracy.

Detection and counting of stamps: I will use a lightweight and fast object detection model (for example, YOLO) so that the system clearly identifies the contours of the stamps on the envelope, counts their number, and classifies them by denomination based on visual features.
…
Data processing and logic: I will write a clean asynchronous script in Python that will accept images, pass them through a processing pipeline, and output a structured result (index, number of stamps, total denomination). In the end, I will document everything.

Why you should choose me:

I have extensive experience in developing complex processing, validation, and data structuring systems in Python.

I know how to build fault-tolerant pipelines with isolated error handling, so the system will not "crash" due to blurred or low-quality photos of envelopes.

I will package the entire project in Docker for easy and quick deployment in your environment.

Please let me know in what format you plan to submit images to the system (a folder with photos, a webcam, or via API)? I am ready to discuss the details and select optimal models for testing!

Show original

Translate

Good day, Dmytro!

I understood the task. I will implement it in Python + EasyOCR (for handwritten text) + OpenCV:
— extracting the index from the envelope (digital and handwritten format)
— detection and classification of stamps (quantity + denomination) using YOLOv8 or CNN classifier

Please clarify: are there test photos of the envelopes? What denominations of stamps — only Ukrainian?

Show original

Translate

Good day!

I can write a similar OCR system in 12 days (during this time I guarantee the full functionality of the system itself as well as all its features).
I have worked on similar projects in Python multiple times (always open to suggestions and ideas).
My portfolio already includes projects related to computer vision, image recognition, data processing, and automation using Python.
I can also offer a convenient system for recognizing indices on envelopes, counting stamps, and determining their denomination.

Show original

Translate

Hello! I will perform OCR for envelopes: recognition of the index (including handwritten — specialized models), detection of stamps with counting quantity and denomination (computer vision + LLM verification). Python, a neat pipeline with manual verification of questionable cases. The price in the bid is conditional — negotiable.

Show original

Translate

Good day. I understand the task: a system that recognizes the postal code from a photo of an envelope (only numbers, sometimes handwritten) and separately counts the stamps, their quantity, and denomination.

Recognition of printed digits of the postal code is reliably solved. Handwritten digits are more complex; here, a training or fine-tuning stage of the model on your actual envelopes is needed, as the quality heavily depends on the handwriting and background. For the stamps, the logic is different: first, detection of the stamps on the envelope as objects, then classification of each by denomination and counting the quantity. For this, a labeled set of stamp samples will also be required.

I roughly see it like this: a separate module for the postal code digits, a separate one for detection and classification of stamps, with a structured result in JSON containing the postal code, a list of stamps, and the total denominations.

To assess more accurately: what volume of envelopes is planned to be processed, and are there already collected examples of photos with handwritten postal codes and various stamps for training? This largely determines how much time will be needed to achieve stable quality.

Show original

Translate

Good day, Dima!
Overall, the task is clear, but to provide an accurate response regarding the deadlines and price, I would like to clarify some questions that arose after analyzing your task.
Please write in private messages – we will discuss the details and your wishes.

Show original

Translate

Good day. I am ready to complete this project as I have extensive experience in app development.

Show original

Translate

Hello! I will implement it using Google Vision API or Tesseract + a custom model for handwriting: recognition of indexes and addresses on envelopes + stamp detection through YOLO with counting the quantity and denomination, all wrapped in a REST API or Telegram bot depending on how you plan to submit the photos as input. 5 days, 10,000 UAH, start after receiving a test sample of envelopes and stamps for model calibration.
I hope for cooperation!

Show original

Translate

Hello!

I have extensive experience in Python development.
I am ready to complete the assigned task.

I suggest we discuss the details in private messages.

Show original

Translate

Good day! I am ready to implement such a system!! Question: what file format will the index and brand be read from??

Show original

Translate

I will develop an OCR system in Python. For text recognition, I will use EasyOCR or Pytesseract, and for labels, I will use OpenCV for area detection + OCR for counting quantity and denomination. Handwritten text is also supported through EasyOCR.

Show original

Translate

Good day, I would be happy to do it. Please write to me privately to discuss the details further.

Show original

Translate

Hello. I specialize in Python development and have practical experience in computer vision projects, automation, and OCR data reading. I will implement a reliable envelope recognition system for you.

Optimal stack for the task:

- OpenCV: frame alignment, contrast, and image preprocessing.
- YOLOv8 / YOLOv11: accurate detection of the index area and finding all stamps on the envelope.
- EasyOCR / PyTorch: recognition of index numbers (even handwritten).
- Template-based classification: to determine the value of stamps (this will bypass the issue of postal stamps that usually interfere with direct text reading).

… To accurately assess the scope of work (cost and development time), I need to familiarize myself with the quality of the input data. Please send 10-15 examples of real photos or scans of envelopes that the algorithm will work with in private messages.

Show original

Translate

Hello! Ready to collaborate! I offer quality and fast work. Write to me.

Show original

Translate

We have a practically ready similar solution for text and object recognition, we can quickly adapt it for envelopes and launch the first working phase, I am available here (:
The estimated cost for the first phase is 65,000 UAH and 21 days for a prototype with index recognition, recipient, number of stamps, and denomination.
There is a nuance - for handwritten text, it is important to see 20-50 real photos or scans, as quality and angle significantly affect accuracy.
Are there examples with the correct answers for the index, recipient, number of stamps, and denomination?
Do we only need to output the result in a table, or is an admin panel or integration with your system required?
A similar class of task example - https://business.ingello.com/vorfahr - automation with applied logic and artificial intelligence.
Another close example - https://business.ingello.com/fractal - complex process automation with artificial intelligence.
About us and the work format on the marketplace - https://systems-fl.ingello.com/ua

Show original

Translate

Hello. I can do it using free libraries or with paid ones. Write to discuss.

Show original

Translate

Hello. For the development of a text and stamp recognition system on postal envelopes, I plan to use a combination of advanced deep neural network architectures. The recognition of postal codes, including handwritten text, will be implemented using adapted models based on Convolutional Recurrent Neural Networks (CRNN) or Transformer-based models, which will ensure high accuracy and robustness to writing variability. Detection and classification of stamps by denomination will be performed using object detectors optimized for speed and reliability. I have successful experience in developing and implementing similar computer vision systems, as well as ready modules to accelerate the implementation of key components. I suggest discussing all implementation details, final budget, and timelines in private messages.

Show original

Translate