Learning is Non-Linear

Some days are Sigmoid. Some days are Relu. Novelty is not inventing new stuff but stitching the right techniques in the right proportion.

A few more I relate to my work style :)

• Persist until you figure things out, digging deeper and deeper into unfamiliar systems until it finally clicks, and you understand how they work.
• Have a broad understanding and creative use of Algos, SQL, LLMs, etc. Teach until you get first principles right
• Have the intellectual curiosity to learn about the many new areas you’ll be taken into. Failures today are arsenal for tomorrow.
• Have a willingness to lean into uncertainty. It's okay if a bad day happens.
• Magnitude improvements and novel capabilities. Keep collecting perspectives.
• You will face unique engineering and AI challenges that will make a meaningful impact. To be authentic, you have to have a lot of scars.
• A focus on outcomes, not time-tracking. Yeah, sometimes easy looks complex, complex looks easy.
• The database is always important. Are not afraid of legacy justice systems and would rather make them work than give up.

Keep Going!!!

Computer Vision License Validation

Business problem: Id verification system(valid or invalid) say driving license as id. How do we go about solving this business problem using Deep learning

Input - License Id Images

Approach

• Feature Definition
• Defining Elements
• Historical data
• Labeling / Annotation

Vision

• Problem #1 - Extract Face images
• Problem #2 - OCR, License Id, Dates, LicenseNumber, Authority
• Problem #3 - Detection for Signature Extracting
• Data Validation - Blurriness - Image Sharpening / Laplacian / Sobel / Canny edge to sharpen images. Non-readable - Far / Validation - Near View

Backend Validation

• API call
• Face Match
• Similarity Score
• Output - Valid License

Keep Exploring!!!

CNN Experiments - Solutions - Building End to End Solutions

CNN Experiments - Solutions - Building End-to-End Solutions

CNN Experiments

• Minimum Exp Without Aug
• Data Aug + CNN Model 
• Data Aug + CNN Model (Deeper Layers) - Few more convolution blocks
• Data Aug + CNN Model (Deeper Layers) - Few more convolution blocks + (Dropouts / Regularizer / Adjusting Learning rate)

To Launch a Product / Build Model things to consider

• Pre-requisites
• Data Collection
• Data Pre-processing and transformation
• Data Imbalances / Data Augmentation 
• Modelling
• Deployment
• Monitoring
• Real-time data training
• Collaborate with Healthcare prof
• Keep updating the model

We have 95% Accuracy, Remaining 5% how do we handle

• Similarity scores
• Ensemble methods
• Human in loop

Keep Exploring!!!

Can ML Solve this Problem ? Vision Problem - How to approach Damage Detection in Mobile Phones ?

How do you approach Damage Detection in Mobile Phones? 

Detecting defects on phones during exchange

Question - Can it be done with ML? 

• Student Answers - DL Vision

Question - Data Prerequisites?

Student Answers

• Physical damage to vision
• Images of the phone from various angles
• Software issues
• System diagnostics
• Images of cracked screens

Question - Model building

Student Answers

• Cnn classification 2 classes
• Damaged, not damaged
• Multiclass - damaged, degrees of damage (so that can identify price negotiation)
• inside parts, maybe images of phone when it is not damaged?

Real-world Way of Solving 

My Recommendation

• Detect Type of Phone, - Flip / Smart Phone
• Brand Detection (OCR)
• Image Similarity (Good Screen vs Similarity score to what you have)
• Line Detection - Count Cracks on Screen
• Segmentation to detect %% of cracked area
• Measure the deformation in the picture
• Yes / NO - Cracks
• Low / Medium / High
• Centre, Lower, Top

This is not a single model for all needs. This has to be based on brands, models, categories, Defect types, Data Collection, Labelling and Phased Adoption.

Keep Exploring!!!

Questions

It's useless to teach people who have no clue how they are going to utilize it. As long as the world runs on certification and theoretical learning no real innovation will come, Before you learn anything question its purpose and outcome

• Have you coded in the past?
• What is the motivation for this course?
• What are the ideas you think you can do with this program?
• Many people drop out when it comes to coding/hackathons? On a scale of 1-10 with strict project / other timelines, Do you have the motivation to perform the training?

Keep Exploring!!!

GenAI based Teaching

Teaching customized with avatars, videos, and images.

Ref - Link

Keep Exploring!!!

Weekend Opinions

On and Off you have to distract yourself when you have problems with development vs sales vs customer expectations.

One interesting link 

The difference between a manager and an expert is a Deep understanding of algos. Sometimes I feel expert, sometimes a consultant, sometimes a manager. The roles and needs keep rotating. i feel comfortable switching Database + Vision not being a full stack developer.

Another Read that hit my mind is link

Concepts need to start with

• Analogy
• Purpose
• Relatable terms
• Mathematical Explanation
• Working examples

We can't learn everything with just maths, formulas, or package and function names. What is missed is blending it and simplifying it. Yes, It is an art to explain in a relatable way. This is the reason we had to spend so many blogs to find one good read :)

Keep Questioning!!!