Python/Django App- Create & Deploy a Computer Vision Model

Explore building a Python/Django computer vision model using CNN and transfer learning to detect car images, check damage status, locate damage, and assess severity to generate a final report.

Explore how CNNs create feature maps with detectors, apply ReLU nonlinearity, pool to reduce size, flatten, and feed into a fully connected classifier to label images.

Create and deploy a computer vision model in Python/Django by training a model, extracting top predictions, building a car category list, and validating whether an image depicts a car.

This lecture demonstrates building a two-class car damage detector using transfer learning on the video 16 base, extracting features, encoding labels, and training a logistic regression classifier.

Train a logistic regression classifier on extracted features to distinguish damaged from whole cars, evaluate with a confusion matrix, and save the pickle model for future use.

Determine car damage location by building a three-class logistic regression model (front, rear, side), training on labeled images, and saving features, labels, and the trained model.

Train a logistic regression model on features and labels from damaged car images (front, rear, and side), save the classifier, and evaluate with a confusion matrix at about 69.9% accuracy.

Load the trained classifier and extract features using the base model to predict car damage location (front, rear, side) for new images in the third check.

Train a three-class classifier that labels car damage as minor, moderate, or severe, using a deep feature extractor, with data preparation, labeling, and saving features, labels, and the trained model.

Learners implement a four-check pipeline to assess car damage severity using a logistic classifier, from car detection to damage location and severity levels, with evaluation notes.

Combine four models for checks into one Python and Django full-stack engine, loading pretrained classifiers in a notebook and deploying with a Django-based front end.

Understand the full stack architecture of the car damage portal, detailing how the frontend and backend integrate using HTML, CSS, JavaScript, and Python Django.

Build the home page of the portal by creating a navigation bar with two links, adding a car image background, and using bootstrap container divs to structure the page.

Create a reusable Django home page using a base template with blocks for a shared navigation bar, then extend it for page content and style with a static CSS layout.

Build portal's second page in the Django app with a navigation bar and an image upload form that triggers the computer vision model and shows results on the third page.

Create the portal's second page with a form to upload an image and email, saving the image to the database via Django models, then render results on the third page.

this lecture builds the portal's second page with an image upload form, configures Django routes and settings, and saves images to a dedicated pick_up_load folder.

Display the uploaded image on the second portal page using the image tag and the uploaded file path, completing the front end. The third page integration follows.

Integrate client and server sides in a Django app to upload images, run a computer vision model for car damage, and display location, severity, and results on a portal.

Integrate client and server side in a Django portal for car damage assessment. Build the third page, enable image upload and processing, and reuse the navigation.

Create a simple Django project with an index page using templates and a basic view. Then learn how to host the page on PythonAnywhere.

Deploy a Django project to the internet by pushing code to GitHub, creating a PythonAnywhere web app, configuring WSGI, and applying migrations.

Build and save a logistic regression image classifier from new features and labels, train on a train/test split, and evaluate with a confusion matrix and accuracy.

Upgrade a Django project to version 3.2.2 using pip, update four files, adjust imports like reverse, remove regex usage, and configure static files and requirements for a computer vision app.