TensorFlow and Convolutional Neural Network, KPR Institute Engineering and Technology, Autonomous Engineering Institution, Coimbatore, India

The Following Projects were assigned to the students. Teams are formed for Execution of projects. Evaluation done for 100 marks. Students gained Practical knowledge using this one credit course from industry experts.
Academic year: 2025-2026

Class: III CSBS              Semester: V

BATCH: 2023-2027

One Credit Course
Code: U21OCB05          Course: TensorFlow and Convolutional Neural  Network

DATE:  10.10.2025 & 11.10.2025      

PROJECTS Assigned to students

1.       House Price Predictor using Keras and TensorFlow

Evaluation Focus

•Use of regression with 2D numerical data

•Model performance visualization (loss curve)

2.       Image-Based Digit Classifier using CNN (MNIST)

Evaluation Focus

•Tensor reshaping and normalization

•Use of Keras Sequential model and Conv2D

3.       Binary Emotion Detector from Text using Keras

Evaluation Focus

•Preprocessing: tokenization, padding

•Binary classification (positive/negative)

4.       AI Stock Market Trend Classifier (Up/Down Prediction)

Evaluation Focus

•Tabular time-series like data

•Use of metrics like accuracy and confusion matrix

5.       Object Detection Lite using Pretrained Models in TensorFlow

Evaluation Focus

•Use of MobileNet or VGG

•Fine-tuning on limited images or test data

Evaluation: 100 Marks

Criteria

Marks

Project Functionality 30

GitHub Maintenance 20

Model Design + Data + Evaluation 30

Viva Questions 20

Total Marks 100

AI Project Evaluation Criteria (100 Marks Total)

Criteria Marks

Evaluation Questions / Guidelines

1. Viva Voce

20 Marks

✅ 4 Questions × 5 marks each:
1. What is the need for AI in the industry?
2. Explain the difference between ML and DL.
3. What are tensors in TensorFlow?
4. How do optimizers help in training deep learning models?

2. Project Implementation & Functionality

30 Marks

✅ Evaluate on:
✔ Model training and prediction working?
✔ Use of TensorFlow/Keras modules
✔ Clear application of ML/DL concepts
✔ Basic UI or notebook-based output

3. GitHub Maintenance

20 Marks

✅ Evaluate on:
✔ Well-structured repo
✔ README file with proper documentation
✔ Code modularity and comments
✔ Commit history showing progress

4. Model Design, Data & Evaluation

30 Marks

✅ Sub-divided as:
• Data Quality & Handling (10 marks):
- Clean, preprocessed data used?
- Normalization, encoding or augmentation?

• Model Selection & Justification (10 marks):
- Appropriate architecture used (e.g., Dense for 2D, CNN for images)
- Justified layer selection, activation, and loss

• Training & Evaluation (10 marks):
- Performance metrics shown (accuracy, loss)
- Any visualization of results?
- Validation or test accuracy discussed?