Scribd
Upload
2 views2 pages

Speech Pfe

The AgriHeal project aims to protect crops from fungal, bacterial, and viral diseases using a mobile application and TinyML technology, specifically with the ESP32 microcontroller. It employs a lightweight MobileNetV2 model for image classification to diagnose plant health and provide care recommendations. This initiative enhances food security and sustainable farming practices by enabling farmers to diagnose plant diseases offline, even in remote areas.

Uploaded by

oufadwa4
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
2 views2 pages

Speech Pfe

The AgriHeal project aims to protect crops from fungal, bacterial, and viral diseases using a mobile application and TinyML technology, specifically with the ESP32 microcontroller. It employs a lightweight MobileNetV2 model for image classification to diagnose plant health and provide care recommendations. This initiative enhances food security and sustainable farming practices by enabling farmers to diagnose plant diseases offline, even in remote areas.

Uploaded by

oufadwa4
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

Each study of fungal infections, bacterial blights, and viral

outbreaks in crops becomes a shield against potential famine,


economic collapse, and environmental devastation. The precision
required to decode how pathogens attack plants, how resistance
develops, and how diseases can be prevented or managed reflects
the extraordinary complexity and beauty of protecting life itself
This work represents far more than academic pursuit; it is the
cornerstone of food security, the guardian of agricultural heritage,
and the foundation upon which sustainable farming practices are
built.
Our project Agriheal is one of the solutions that can help farmers to
protect life in our plant ,it has two principale parts: smart farme
using mobile application and systeme detection using tinyml
especially ESP32.
To bring this project to live we did an organised detailled work
starting by choosing the suitable dataset,train a model and deploye
it.
Choosing a CNN model was also a crusial step in building our
project, we needed a fast, lightweight,depthwise separable
convolutions model,and mobileNetV2 was the best choice, it is
preferred for its lightweight architecture, strong balance between
speed and accuracy, Its compatibility with TensorFlow Lite and
support for quantization allow smooth deployment on mobile apps
and microcontrollers like the ESP32.
***Using TensorFlow and Keras to train a lightweight image classification model using
MobileNetV1 with a width multiplier of 0.25, making it highly efficient for embedded systems.
The CIFAR-10 dataset, consisting of 10 image classes, is loaded, resized to 96×96 pixels,
normalized, and prepared using the TensorFlow tf.data API for optimized input pipelines. The
MobileNetV1 model, pretrained on ImageNet, is initially used as a frozen feature extractor, and a
classification head is added for training. After an initial training phase, the last 20 layers of the
MobileNet backbone are unfrozen for fine-tuning with a lower learning rate to enhance accuracy.
The trained model is then saved in the Keras format. To enable deployment on microcontrollers
such as the ESP32,******

As a an engine for this application we choose flask ,it a simple easy and fast framwork that
garantie a smoth communication with our models.
Scanning page allows the user to choose bettwen Scan LIve where he can take a pic of his plant
and send it to the backend for the state detection either healthy or affected, or he can uplaod
directely his pic ,
Scan Result screen determine the name of the plant , it’s state either healthy or affetcted and
the disease name other than more additionnal information like the pathogene that cause that
diases , some special tips and recommandations care , prevetion TIPS, from the specific disease
IA assistant page is a small guid for the user where he can ask for more informations and get
more knowledge about the disease, how to care with his plant ect ...
******* Our objectif was to carete an offline model with vast informations that can integrate in
The llama-cpp-python project
our mobile app , and to realise this we used
provides a simple Python interface to run LLaMA models via
llama.cpp which was an ideal choice for our context to use this
interface we created a virtual Python environment in .x64 Native
Tools Command Prompt for VS 2022 to manage dependencies
cleanly. Then, install the optimized AVX2 version using a custom
index URL for better CPU support. Download a .gguf model like
gemma:2b-7b-chat.Q4_K_M.gguf from Hugging Face (e.g.,
TheBloke repo). This setup allows us to run powerful LLMs locally
without needing a GPU or cloud service.and after this we integrate
using flask .*******

The Plant Disease Detection project combines AI and TinyML to deliver fast, offline,
and accurate crop disease diagnosis.
Through a mobile app and ESP32-based system, it empowers farmers even in remote
areas without internet.
“Technology will not replace farmers, but farmers who use technology will replace
those who don’t.”
AgriHeal shows that low-cost, intelligent tools can revolutionize agriculture and
support a more sustainable future.

You might also like