I need your help in my project plz

AI Discussions
,
1

I have an important project and I will make a model that recognizes the parts of the car and I don’t know where to start from and I don’t know how to get a research paper to help me so can I help
Thank you

1 Like
2

What machine learning courses have you attended?

1 Like
3

Machine learning specialization with Deep learning.ai

1 Like
4

If you want to recognize objects in images, you will need some deep learning techniques (for example, convolutional neural networks).

That’s covered in Course 4 of the Deep Learning Specialization.

1 Like
5

And I will use computer vision also but I don’t know how to use it if u have research paper to help me, that would be enough.

6

Sorry, I don’t have an references for that.
I recommend you attend a course and learn the techniques.

2 Likes
7

Explore computer vision model named ‘RESNET’ or ‘VGG16’ or such similar models and see which one gives good results on your usecase.
Build dataset of images of car parts (example). Do this by making folder and naming it with that particular name.
Finetune Resnet or any such computer vision model using your finetuned dataset. For this,
save the finetuned model and use it further based on your use case

2 Likes
8

finetuning is also called transfer learning where we retrain certain layers of the network to make the network recognize data that we have trained it on

1 Like
9

@kali.mane26 @YF-ZG I think Tom has a good suggestion-- It can be difficult to use these models properly if you don’t know some of the theory of how they work.

My greater concern though, honestly, is where are they going to find a pre-marked dataset with the parts of cars ? I mean it might exist, but, if not, they have a massive data labeling task on their hands to take care of.

3 Likes
10

This is an excellent point.

1 Like
11

Data science projects involve preparing our own dataset to train networks for production level usecases…actual projects are long and involve this data preparatory phase…

1 Like
12

I can help with the theory and the code around computer vision models … let me know…my email–>maneesha.vinayak@gmail.com

2 Likes
13

Try the Tensorflow Developer in Practice by Deep Learning AI and Go through the first 2 modules. You will need a nice dataset for car parts. But you will be able to build a model. Then you can use flask or tensorflow serve to use it in your application.

2 Likes
14

There is a short course on here for prompt engineering with image models: Prompt Engineering for Vision Models - DeepLearning.AI

You will learn to use off the shelf models for segmentation and object detection. Models like CLIP likely already know what car parts look like. It’s 2h and beginner level. You may not need to the deeper learning and work training your own models.

2 Likes
16

If you are looking for parts of a specific car, perhaps you want to scrape a dealer’s manual for that car. If you want car parts in general, get the inventory from an Auto Parts store. Is that what you mean?

1 Like
17

If you have a link for the CLIP model you referenced, could you post it?

18

You can install CLIP with pip, instructions in their repo here: GitHub - openai/CLIP: CLIP (Contrastive Language-Image Pretraining), Predict the most relevant text snippet given an image

If you want to use it for classification, you can get the text embedding of each of your classes, get the embedding of the image, and then find the closest by cosine similarity.

(You might also just try asking gpt-4o with an image upload…)