OCR: NLP, Geometric, GAN
$1500-3000 USD
착불
1. Description
We have a project that reads documents and we need to quickly get it off the ground and get it moving so we can go live. The number of formats is 500 million different designs on the documents. So it cannot be a template based solution.
Please note this is not a long bench with months, we are talking about 1-2 weeks at most.
We have not started mapping the labels but we have Tesseract v5 running and it stores it in the DB for sure.
So, we now need to do some text/keyword value mapping.
- it is focus on the detection of strings.
- SPACY, NLTK, Gensim is on the table, which one is best and fastest?
- Geometric detection is then the best way.
- Graph Neural Network is an alternative to enhance the cross-checks.
The values we must map is in the attached file.
Milestones and some estimated time to do it.
MS1 OCR and parser
MS2, 12 labels, easy
MS3, 20 Labels, cross-check with geometric and tabula
MS4, 5 labels
MS5, 5 labels, we might skip them
We work here and now, not in 1,2 or 3 weeks. The platform must be launched now.
We also need to be able to compile the python code and run it under W10 or MAC.
This process is handled by another developer.
We will when we have more documents be able to do ML/DL/NN also to complement it.
For clarification: I do not expect 100% in precision the first day, but you must know what you are doing.
You are the expert and you should be able to advice me what the best way is. I have a wealth of experience, but you get humle from knowledge how little you know.
The drive forward is of essence. Do not suggest long term projects here. Now it is time to market so please consider that.
2. Skills:
Python, NLP, geometric detection, label parsing
3. Time frame:
1-2 weeks approximately.
4. Budget
$1000-$3000
5. Number of developers
1-2
6. Code stack
Python, must be possible to make it compiled python (our other developer will do that)
I am looking forward to your proposal.
프로젝트 ID: #28954775
프로젝트 소개
이 일자리에 대한 프리랜서 18 명의 평균 입찰가: $2320
Hi, Greetings! ✅ I have checked and gone through the project details and I am very skilled in it. ✅Completed Time: In project deadline We have worked on 640 + Projects. I have 6 + years of the experience in sa 기타
Hi, I hope you are doing fine. I have almost 10 years of experience in machine learning algorithms. I can implement various types of artificial intelligence algorithms including yours with Matlab, Python and etc. I hav 기타
Hi, Robert. I just read your posting. I have rich experience in OCR using python for 5 years. so I think we can help you about this issue. Let's discuss more in chat Best Regards Elite Lancers Team
What exactly you are trying to say? Can you please explain in more detailed manner? It seems all puzzled sentences to me. What exactly are you trying to build? And whom are you referring by "We"? Its all confusing desc 기타
Hello, I am a PhD candidate in Electrical Engineering (control systems). I'm expert in MATLAB and python programming and by strong mathematical and statistical background, have good flexibility for solve your project 기타
Hello there. We will build a machine learning platform so that new variants of documents can be passed through it. Kindly contact me and lets discuss more.
Hi, Thanks for your job posting. I have read your job description very carefully and understood what you want. In my opinion, I personally prefer using NLTK for the pre-processing of natural text (i.e., gathering, wra 기타
Hello , i am a full stack developer with 6 year experience in ML, data mining , geospatial , python etc, I have seen your requirement and I can do it , please check my profile https://www.freelancer.in/u/arjun366333, w 기타
We have experience in Software development, Machine learning, OCR, Image processing. Kindly message me to discuss the project details.
***** I Love To work With You ***** Hi, nice to meet you? I’ve checked your requirements very carefully, so I understood what you want now. I am an expert with many years of experience in developing deep learning tasks 기타
Hi, As you remember we have talked details yesterday. Milestone 1 will be training a NER model with a small invoice dataset. Then if results are satifying, we will continue with other milestones.
