Computer Vision Research

• Face Recognition / Unlocking: Using the front camera to identify your face and unlock the phone.
• AR Filters / Effects: Detecting facial landmarks to apply augmented reality effects (like Snapchat or Instagram filters).
• Object Detection in Photos: Identifying objects in images for tagging, searching, or automatic categorization (e.g., Google Photos).

Objects identified:

• Laptop
• Smartphone
• Notebook
• Pen
• Water bottle
• Mouse
• Desk lamp

1. Image acquisition: The camera captures the desk photo.
2. Preprocessing: Adjust lighting, normalize colors, resize the image.
3. Feature extraction: Detect shapes, edges, and textures of objects.
4. Object detection / recognition: Classify objects using trained models.
5. Labeling / output generation: Assign names to the detected objects.

So roughly 5 major steps are involved.

• Lack of visual input: ChatGPT cannot literally “see” images, so it relies on descriptions instead.
• Ambiguity in objects: Objects partially hidden or overlapping can be misidentified.
• Context interpretation: Determining the function or purpose of items is difficult.

Asked to highlight water bottle:

Challenges:

• ChatGPT cannot literally draw or circle images.
• Cannot verify accuracy visually.

• Self-driving cars (deciding to brake or turn based on detected objects).
• Automated quality inspection in factories (accept/reject products).
• Security cameras with AI (alerting security personnel for unusual activity).

Reference: medium.com

• Surveillance & facial recognition: Can lead to privacy violations and misuse.
• Deepfake generation: Can produce misleading videos/images for manipulation or fraud.
• Autonomous weapons targeting systems: High-risk decisions with potential safety hazards.
• AI-based hiring systems: Can perpetuate bias if training data is not carefully managed.