SORTING TOMATOES BASED ON SIZE AND COLOR USING TENSORFLOW AND ARDUINO

Подписаться 278

Просмотров 44 тыс.

50% 1

*** (maximum support duration is up to date 31/12/2019)
A part of abstract:
Thesis uses Raspberry Pi 3 as the main controller to take images of fruits with a camera, forwards the image to a TensorFlow neural network running on a Linux server to perform a more detailed classification of tomatoes has different color, shape, and weight.
We have used the idea from Makoto Koike's project to created our machine.
Makoto Koike's project: cloud.google.c...
We define 5 types of tomatoes:
Type 1: perimeter greater than 20 cm and ripe red tomatoes completely.
Type 2: perimeter greater than 20 cm and not yet ripe red tomatoes (you know about 70% process to become ripe red)
Type 3: perimeter smaller than15 cm and ripe red tomatoes completely.
Type 4: perimeter smaller than 15 cm and not yet ripe red tomatoes.
Type 5: green tomatoes and things that the machine don't know.
Accuracy rate: 70%.
Materials: Conveyor + Combo Raspberry B+ + Camera + Relay 5V + Power Suppy 12V 5A + Arduino Uno R3 + Servo 9G + ESP8266v1 + Loadcell 5kg + Loadcell HX711 + Mica + Data

Опубликовано:

16 окт 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 50

@yencytolentino4299 2 года назад

Hi there, can I ask the brand or at least the specs for the cameras used in this project? Also what is the minimum amount of camera used in this machine? Thank you!

@maysoonahmed7410 3 года назад

I want to ask you about the data set you used and the recommendations of your thesis?

@johnlao1469 4 года назад

Hi. Is there any chance that i can see your source code? We have the same concept on our project instead of tomatoes we used pineapples.

@plavali_znaem 4 месяца назад

did you find the source code, or were you able to finish your project and maybe could share your code?

@andinelinggapradhana9055 11 месяцев назад

Can you share a datsets to learn on college project?

@AHGaming0720 Год назад

can i use arduino in project.

@nguyenanhquoc7704 10 месяцев назад

Anh ơi cho em tham khảo code được không ạ? Em đang không biết xử lý đoạn sẻ

@nguyenduykhanh3781 5 лет назад

thuật toán điều khiển chỗ servo phân loại sao vậy ạ? Em cũng đang làm đề tài gần giống như vậy mà em chưa biết viết thuật toán điều khiển sao nữa? Em chỉ có thể phân loại từng trái( trái này phân loại xong rồi mới tới trái tiếp theo được) chứ không chạy liện tục như thế kia được ? Mong anh giúp đỡ

@minhhoangle536 5 лет назад

Bạn gửi mail đến một trong các email ở description nhé. (Không biết sao tự mất reply rồi.)

@nguyenduykhanh3781 5 лет назад

@@minhhoangle536 cảm ơn anh em nhận được email rồi ạ!

@satishbvkumar9528 Год назад

@@minhhoangle536 pls send me ur email id

@johnlao1469 4 года назад

Hi. Using Image classification from the guide in tensorflow will classify also the size of the object? like this tomato is big small normal.

@emerjay6754 4 года назад

What if I'm just going to sort out objects according to its size? Like plastic bottles. Into a 3 different ranges of sizes

@raymurutempest2404 2 года назад

Can I asked what kind of software did you use?

@minhhoangle536 2 года назад

"Tensorflow" for detection.

@lqcgaming2483 4 года назад

làm sao để camera truyền tín hiệu cho servo vậy ạ. Cho em xin code với

@Futureispeace 4 года назад

Which types motor you used?servo motor?

@Adrian20Xx1 5 лет назад

Excuse me, does the data normalization will effect the color grading ? And did you use the Image Classifier from TensorFlow's guide. Thanks in advance.

@minhhoangle536 5 лет назад

Hi, 1st question: yes, data normalization will affect color grading. Because, in the study room conditions, data normalization will make your thesis more beautiful, I mean its easy to demo and you must select your samples carefully. But, in real-life conditions, data normalization will low your accuracy, because of the appearance of variations. In my thesis, I didn't use data normalization because it is the mission of the tranning process. But If you can standardize the database before training, you can speed up the training process. 2nd question: yes, just a little part. I changed a lot of code, but the theory is based on the Image Classifier from TensorFlow's guide.

@Adrian20Xx1 5 лет назад

Thank you for replying. I based mine from the tensorflow's image classification guide. During the training, it has 1.00 val accuracy, but when i tried to load the saved model of that trained model it doesn't predict correctly, i always get [[0.]]

@minhhoangle536 5 лет назад

How much is the loss after training? How big your data use to train?

@Adrian20Xx1 5 лет назад

Quite small, I have 50 for each red and green. and 20 each for validation. The loss is nearing 0, 2.9e6 something like that.

@minhhoangle536 5 лет назад

Yeah, it is really a small database for training a new model. Try to get more data and see how it affects your system. If you need, I can send a part of my old data that I used for my system.