Scribd
Upload
5 views2 pages

GeoSpy AI Computer Vision Evaluation

The document outlines a technical evaluation for a Computer Vision Engineer position at GeoSpy AI, focusing on designing and implementing an AI system to predict geographical regions from images. It consists of three phases: system design and model selection, implementation of the core model, and evaluation with error analysis, with specific deliverables for each phase. The evaluation criteria emphasize model design, implementation completeness, and error analysis, with a strict prohibition on the use of AI tools during the task.

Uploaded by

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

GeoSpy AI Computer Vision Evaluation

The document outlines a technical evaluation for a Computer Vision Engineer position at GeoSpy AI, focusing on designing and implementing an AI system to predict geographical regions from images. It consists of three phases: system design and model selection, implementation of the core model, and evaluation with error analysis, with specific deliverables for each phase. The evaluation criteria emphasize model design, implementation completeness, and error analysis, with a strict prohibition on the use of AI tools during the task.

Uploaded by

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

GeoSpy AI - Computer Vision Engineer Technical Evaluation

Objective:
Your task is to design and implement a prototype AI system that predicts the geographical
region of an image using computer vision and deep learning techniques. This exercise will
evaluate your ability to work with image feature extraction, geolocation AI, and model
optimization.

Phase 1: System Design & Model Selection


• Design a computer vision pipeline that processes images and extracts location-based
features.
• Compare two potential models:
o CLIP (Contrastive Learning for Image & Text Matching)
o Vision Transformers (ViT-based Feature Extractors)
• Justify your choice based on accuracy, efficiency, and scalability.

Deliverable: A one-page document explaining your approach and model choice.

Phase 2: Implementation - Core Model


• Implement a prototype that can:
1. Extract image features using a deep learning model (e.g., ViT, CLIP, or ResNet).
2. Predict the most likely geographical region from an image.
3. Evaluate the model using test images.

Deliverable: A Python script or Jupyter Notebook that takes an input image and outputs the
predicted region.
Phase 3: Evaluation & Error Analysis
• Test your model with at least 5 unseen images.
• Analyze mispredictions and identify key failure cases (e.g., occlusions, urban vs. rural
settings).
• Propose one improvement that could enhance the model's accuracy.

Deliverable: A brief report summarizing:


• Model performance.
• Error analysis.
• Potential improvements.

Evaluation Criteria:
Metric Weight (%)
Model Design & Justification 30%
Implementation Completeness 50%
Error Analysis & Improvements 20%

Submission Format:
• A ZIP file containing:
o Design document (PDF)
o Code (Python script or Jupyter Notebook)
o Evaluation report (PDF)

Deadline: Sunday 9th February, 2025 EoD

Use of AI tools like Copilot, ChatGPT, or similar for this task is strictly prohibited. We employ sophisticated
detection methods to identify any use of such tools. Violation of this policy will result in immediate termination
of the process. Please rely solely on your own skills and the provided resources. Contact us for any
clarifications.

Good luck! We look forward to reviewing your solution.

You might also like