Modeling and Simulation of Real Time Object Detection in Autonomous Vehicles
Implementation Plan:
—————————
Step 1: Initially, we collect and load road scene images from the KITTI Dataset
Step 2: Then, we pre process the images using perform image resizing, image augmentation and normalization techniques.
Step 3: Then, we initialize a lightweight YOLOv8 model for training based on collected data.
Step 4: Next, we apply knowledge distillation where the model learns from the soft predictions and feature outputs.
Step 5: Next, we train a YOLOv8 model based on loaded images to detect objects using object detection features.
Step 6: Then, we compute dynamic contribution scores for each training round using loss improvement, confidence score, and gradient change.
Step 7: Next, we optimize the model using the FedAvg algorithm to aggregate the high-contribution model weights and reduce latency .
Step 8: Finally, we evaluate and plot the following performance metrics:
8.1: Number of epochs vs. Accuracy (%)
8.2: Number of epochs vs. Precision (%)
8.3: Number of epochs vs. Recall (%)
8.4: Number of epochs vs. F1-score (%)
8.5: Number of epochs vs. Latency (ms)
8.6: Number of epochs vs. Communication Cost (MB)
Software Requirements:
—————————–
1. Development Tool: Python 3.11.x or above version
2. Operating System: Windows 10 (64-bit) or above
Dataset Link:
—————–
Link:- https://www.kaggle.com/datasets/ibrahimalobaid/kitte-dataset
Note:
——-
1) If the proposed plan does not fully align with your requirements, please provide all necessary details—including steps, parameters, models, and expected outcomes—in advance. Kindly ensure that any missing configurations or specifications are clearly outlined in the plan before confirming.
2) If there’s no built-in solution for what the project needs, we can always turn to reference models, customize our own, different math models or write the code ourselves to fulfil the process.
3) If the plan satisfies your requirement, Please confirm with us.
4) Project based on Simulation only.
