Find Jobs
Hire Freelancers

Image Recognition to match images from a database of distinctly generated images

$1500-3000 USD

Lezárt
Kiadva ekkor: több mint 5 évvel ezelőtt

$1500-3000 USD

Teljesítéskor fizetve
We are a young start-up company founded by a group of scientists. In this project, we are seeking a freelancer who will be able to pull off a small Computer Vision project in Python. In the scope this job; 1-) Develop a small tool that generates distinct images similar to Jdenticon ([login to view URL]) for id numbers like FL1749593, AZ945430, TX9506813 etc. 2-) Develop an image recognition/matching algorithm that recognizes an image in a scene and matches with the database of images. 3-) Implement the algorithm and develop a small tool. The tool should output the corresponding ID number when a scene image is fed into the tool. Important Notes: * Project will include milestone payments and payment will be made after approval of each milestone. ** The source code will be in Python language (the code can tunnel an API available in other languages). The code will be properly commented with clear explanation. *** The image recognition should be robust to color, therefore while generating images the color should not be used as hash value (tool should not generate the same shape in the same color) **** Some example images that involves the database images will be provided later in the project for testing of the tool. (The freelancer can should start with own images) ***** Additional information will be provided during private discussions about the project.
Projektazonosító: 17581300

A projektről

19 ajánlat
Távolról teljesíthető projekt
Aktiválva: 6 évvel ezelőtt

Szeretne pénzt keresni?

A Freelancer oldalán történő árajánlatadás előnyei

Határozzon meg költségvetést és időkeretet
Kapja meg fizetését a munkáért
Vázolja ajánlatát
Ingyen regisztrálhat és adhat árajánlatot munkákra
19 szabadúszó adott átlagosan $2 461 USD összegű árajánlatot erre a munkára
Felhasználó avatár
Hello I have more than 10 years experience in machine learning and algorithms and have worked on many projects in these fields. Regarding this project actually there are many variation of identicon algorithms which could be used. Major steps are generating a digests for the entered data, dividing the outputted image into a set of cells and filling each sell using background and foreground colors generated based in the digests. the main challenges are related to two points: limitation related to the number of possible colors, noisy, and limitations related to the dimensions of the outputted images. but in your case i think we can overcome these challenges since you focus on a specific case (codes like FL1749593, AZ945430, TX9506813 etc), we can design the algo in way where we can generate distinct images. I have one question please to be sure that i understood exactly what you need, you said "recognizes an image in a scene and matches with the database of images." could you please give me more details about what you mean when you said "scene", do you mean an image where there is a "identicons" in it? Thanks
$2 500 USD 30 napon belül
5,0 (40 értékelés)
7,2
7,2
Felhasználó avatár
check my profile and you will speak with me .
$2 000 USD 25 napon belül
5,0 (57 értékelés)
6,3
6,3
Felhasználó avatár
Hi, 1-) How are you going to generate distinct images ? How does hash visualization work ? We can generate distinct images through random art algorithm which is a form of hash visualization. we will be able to generate distinct images as the seed of the algorithm will be distinct and we can generate images through that. 2-) How is your algorithm going to match an image from a large database of similar images ? we can use opencv for pixel based image recognition and will use singular value decomposition which will also handle like problems of image scaling, image rotation, search time 3-) What are your challenges ? the biggest challenge is image registration which can effect the recognition process. I have a lot of experience in Computer Vision applications / image and video processing in mobile and windows applications. My Team experience includes: -Port Based container Damage detection Software: detecting damaged container and keeping record and report of containers movement, -Vehicle counting in the particular category they belong to on a high speed road, -Number plate recognition using OCR algorithm of openCV, -Face detection and recognition for multiple applications. -Body measurement and detection of changes in body -3D reconstruction from 2D multifocus images In short We have all the experience needed to complete the project with maximum accuracy. I will be waiting for your kind response. Regards
$2 500 USD 30 napon belül
4,9 (12 értékelés)
6,1
6,1
Felhasználó avatár
Hello Sir! I read your project carefully and I am interested in your project. I am a image processing expert and I have rich experience in this field. I can build your project using c++ & python. I have mastered in opencv & tensorflow. If you give me a chance, I will give you happiness. Thank you. Best Regards from Yan.
$2 500 USD 30 napon belül
5,0 (13 értékelés)
5,6
5,6
Felhasználó avatár
Hello, Greetings of the day.!! please ping me i can tell u stepwise approach to achieve this image identification form A db. Your project attracted my attention at first glance, because I've really rich experience in Machine Learning & Python Programming. I have worked on several similar projects before! I'm really confident about your project, and very eager to join your project. I am having 3+ year of experience in Data Science using tools like Python, R, Spark. Worked on multiple projects in various domains like life sciences, CPG, Insurance. I have completed Master of Technology in Machine Learning. If we have a chance to cooperate, I'll do my best to provide wonderful result. Work Areas - Image Processing, Text Mining, Supervised and Unsupervised Machine learning problem, Voice Analytics. Thanks, Analix D
$2 444 USD 100 napon belül
4,6 (60 értékelés)
5,8
5,8
Felhasználó avatár
Hi! A very interesting project. I understand the goal is to use the algorithms developed in this project for road sign tagging and recognition? a) Based on the examples given, we need to map the hash to an encoding describing the visual identifier. E.g. for the Identicon example we would simply map the hash (bitwise) to a list of filled/unfilled values for each of the blocks in the image (which is vertically symmetric). A more complex language (similar to Jdenticon) would include bits e.g. for vertical/radial symmetry, basic shape (circle/triangle) per image block, or color saturation. b) For the reverse procedure (visual identifier to hash), we would generate a large number of distorted/rotated/shaded/etc. images to train a neural network (a CNN) with. This could include a non-machine-learning part where we manually decode the output of the nn. c) There are many good Python machine-learning libaries, I have some experience with TensorFlow. One challenge, if I have understood the requirements correctly, would be to make the image generation and recognition robust enough so that real scene images are recognized correcly too. Another challenge is error-detection/correction. The visual-identifier space needs to be "sparse" enough to prevent mis-recognition. Thank you for reading, best regards, Panagiotis
$2 222 USD 40 napon belül
5,0 (6 értékelés)
4,0
4,0
Felhasználó avatár
Hi 1-) How are you going to generate distinct images ? How does hash visualization work ? We can make use of QR code library to generate QR codes instead of distinct images, generating QR code is not only easy process but you may use it with other apps too to detect ID. 2-) How is your algorithm going to match an image from a large database of similar images ? We will be building Deep Learning Model with the help of neural networks where model will be trained on existing images and if image detected it will return label of image which will be searched in database. 3-) Which methods are you going to use ? What are your challenges ? We will be making use of convolution neural networks which are popular for computer vision tasks now a days. Main challenge is data set , can you please mention how many images you have so far i.e. training data. I have been working on Machine Learning algorithms like Regression, SVM, Decsion Trees and Deep Learning RNN, CNN and GANs Lets discuss
$3 000 USD 40 napon belül
4,6 (1 értékelés)
4,1
4,1
Felhasználó avatár
Hello , I am a Certified Software Developer with 7+ years of experience in programming , Algorithm, Image Processing, Machine Learning, Python, Software Development and Application development . I have reviewed your project description and I would like to develop a tool that generates distinct images like Jdenticon and develop an image recognition algorithm that recognizes an image in a scene and matches with the database of images and all other features as per your requirements . I have developed Software's and applications from design to coding and full implementation . I have consistently produced terrific code within customer-set deliverables time frames. I am responsible for creating, developing, benchmarking, verifying and testing of various applications which are mainly based upon online interaction and to build and design the module according to the specification of clients and deliver it in proper deadline. I would appreciate the opportunity to meet with you further talk about my previous experience. Best Regards.
$1 500 USD 10 napon belül
0,0 (0 értékelés)
0,0
0,0
Felhasználó avatár
Identicons are not designed to be unambiguous, so I assume you mean only visual resemblance. The symmetric style of identicons is wasteful for ID coding, a lot of bits could be fit on the same area. So I propose to use a method based on QR codes by changing the squares to triangles. If you insist to use symmetric shapes, it means the patterns will be larger. About the recognition phase. I agree with you on color based encoding. There are a lot of problems with white balance out there. I propose therefore that colors will be identicon style: deterministic but ambiguous, using only darker than 50% gray colors. Grayscale images should be decodable. For first milestone, I would say experimental image generator software with theoretical possibility of recognition without recognizer program. This is useful to clarify and change the style of the images, makes us able to finalize the features of the images. For the second milestone I can make a recognizer. This is a standalone program reading an image from file, and you can use the source code to integrate wherever you want to. Test images would include photos of printed ones. We create a test set of at least 100 images, and I need to provide a solution which works on >90% of them for the milestone. That's what I think is reasonable to offer in a few weeks, later more. About your thoughts on QR codes in the pdf: I do not agree with you on most issues, and I think I can convince you in chat, I have run out of characters here.
$2 500 USD 30 napon belül
0,0 (0 értékelés)
0,0
0,0
Felhasználó avatár
Hi, I read through the description. I have developed a small POC on my end to test whether this is possible. 1. For generating distinct images (which are color agnostic, I'm going to assume internal processing is done in monochrome/grayscale) which do not wear out easily in physical conditions or obstacles, it is better to use simpler shapes like plain circles in the renderings.A 4x4 grid of circles will give us an effective 2^16 combinations for IDs in the database. These dots/circles can be easily decoded in low resolution as well because the existing circle detection algorithms are very robust to obstructions and wear/tear. We can then decode the grid into a DB Id, from which we can pull the information. 2. The generator will generate the grid surrounded by the 3 position markers and an alignment marker as is usually done in qr code generation. This is also advantageous as QR code boundary detection algorithms keep getting better and hence we can exploit that advantage for boundary detection. 3. Since a grid can be equivalent of a binary representation, it is easy to decode and match to ID in DB. 4. The biggest challenge get boundary detection and perspective correction accurate. 5. I am also aspiring to do a computer vision/machine learning Masters in an Ivy League institute, hence I feel this project really motivates me to accomplish it at the soonest. Please message me to discuss further about the tools i've currently used for the POC.
$2 777 USD 20 napon belül
0,0 (0 értékelés)
0,0
0,0

Az ügyfélről

UNITED STATES zászlója
ORLANDO, United States
0,0
0
Fizetési mód hitelesítve
Tagság kezdete: nov. 13, 2016

Ügyfél-hitelesítés

Köszönjük! E-mailben elküldtük a linket, melyen átveheti ajándék egyenlegét.
E-mailje elküldése során valami hiba történt. Kérjük, próbálja újra.
Regisztrált Felhasználók Összes Közzétett Munka
Freelancer ® is a registered Trademark of Freelancer Technology Pty Limited (ACN 142 189 759)
Copyright © 2024 Freelancer Technology Pty Limited (ACN 142 189 759)
Előnézet betöltése
Hozzáférést adott a helymeghatározáshoz.
Belépési munkamenete lejárt, és kijelentkeztettük. Kérjük, lépjen be újra.