Day #205 - Short ML Talks - Agriculture

Talk #1 - WACV18: DeepWheat: Estimating Phenotypic Traits from Crop Images with Deep Learning

Key Lessons

• Working on Aerial Image

• Automate Inspection for Bio-Mass
• Plant counting high rise picture
• Biomass Estimations

• Non-Linear Regression
• CNN for both
• Plant Counting, Leaf Counting 

Talk #2 - Implementation of Deep Learning in Agriculture Crop Identification

Key Lessons

• Identification of crop from satellite Images with crop based Deep Learning Models
• Google map satellite imagery
• Extract Image tiles and label them
• Vegetation Index with Chlorophyl Content
• Multispectral Imagery for Crop Health Analysis

Talk #3 - WACV18: Recognition of Pollen-bearing Bees from Video using Convolutional Neural Network

Key Lessons

• Ecological aspects Bees Disappearing
• Dataset setup
• Identification and Activity Analysis

Next Talks

WACV18: An Animal Detection Pipeline for Identification
Face Liveness Detection Based on Perceptual Image Quality Assessment Features 
WACV18: Crowd Counting With Minimal Data Using Generative Adversarial Networks
WACV18: Workshop: CDBR: Cross-Domain Biometric Recognition Overview
WACV18: Workshop: CV-AAL: Calorific Expenditure Estimation Using Deep Convolutional Network Features
WACV18: Workshop: CV-AAL: Assessing Pain Levels From Videos Using Temporal Convolutional Networks

Happy Mastering DL!!!

Day #204 - MIT Self-Driving Cars: State of the Art (2019)

Key Lessons

• Improve Access to mobility, increase efficiency, save lives

• Reduce stress / emotion in travel
• Save Lives / prevent crash, Every 23 seconds someone dies in Auto-Crash

• Waymo Autonomous 1 Million miles, 1 billion miles reached by Tesla Semi-Autonomous
• Largest Deployment of NN in the world that has direct impact on human life
• Image Classification on Imagenet -> Human Life
• One pedestrian fatality in Uber, Third fatality in Tesla
• Autopilot is driven on particular kinds of roads
• 3 Fatalities is not a large number to make significant conclusions
• Many Trial runs of autonomous tests carried out
• Predictions - commercially viable at scale
• 2020 onwards you will see them hitting on road

• Levels of Autonomy
• Up to 3 Drivers responsible
• > 3 Car handles

• Semi Autonomous - Autonomous transition of vehicles

• Higher goals of reducing emissions
• Tunnels / Connected Train setup with cars

• Personalised flying cars
• Split second decision in limited visibility conditions/weather/ non-verbal pedestrian communications 

• Radar / Sensors for Obstacle detection and obstacle avoidance
• LIDAR expensive, Higher resolution, 360-degree visibility

• Fail Safe with both approach, economical approach
• Fully Autonomous - Waymo

Happy Mastering DL!!!

Build a world class personalized Education System

This post is my personal opinion. There are three types of knowledge in my perspective

1. Fundamentals - The basics of maths, physics, chemistry that underlines every concept we relate to
2. Technology - The innovations/discovery that runs in the technology that is built on the fundamentals
3. Experimentation - The gaps that connect the Fundamentals and Technology with meaningful experiments

I do not believe in Multiple Choice Questions, paper-based expertise. They are good but every method of teaching has their own drawbacks. Outside our regular education, we need to have a constant learning setup, the mindset to experiment, stay curious and generate more ideas. It takes a lot of learning, gathering new perspectives to be creative and passionate about things we do. A personalised education means able to access student is good in fundamentals, theoretical concepts and build the necessary bridge to make the next steps concrete that allows him to build his expertise. We have a disconnected system where we are not encouraged to connect the dots and build a larger system. Education ends when the grades are out.

With such a large pool of students why we are not able to get innovation at grassroots. The system that encourages questions, experimenting and creativity only can lead to innovation. With AI I believe those days are not very far. Access to quality educational content, the ability to provide personalised education is just a few years away.

I believe the intelligence, passion and curiosity to solve will be 10X more with all these resources in place to build a more intelligent, passionate and curious learning mindset.

Keep Learning!!!

Day #203 - Multiple Short ML Tech Talks

Read 1 - WACV18: CSVideoNet: A Real-time End-to-end Learning Framework for High-frame-rate 

Key Summary

• High Transmission & Large Storage Space
• Energy Efficiency is a concern
• Compressive Sensing (sense data from compressed form)
• Motion synthesizing LSTM
• RNN to extrapolate motion 
• Pretrain using LSTM

Read - 2 - WACV18: Recommending Outfit from Personal Closet

Key Summary

• Outfit recommendation
• Dataset created
• Positive and Negative Samples
• Good outfit - Well coordinated / selected
• Resnet50 to extract feature vector
• Positive probability as score
• Binary Classification problem
• Beam Search

Read 3 - WACV18: FARSA: Fully Automated Roadway Safety Assessment

Key Summary

• Deals with low level attributes on road
• Build a grading based on positive / negative samples
• Scores for pedestrians, cyclists
• Panaromic view, labelling
• CNN to directly label
• Image -> VGG -> Star Rating
• Predict Road Level Attributes

Some of Projects Applicable for India

• Automated road quality assessment (Automatically identify and report / Automate inspection)
• Automated unmanned Railway Crossing gates (Use Audio + Light to eliminate manual tasks)
• Automated CCTV monitoring and person identification / Alert for better security monitoring 

Rather than fixing the problem, we need to nail out the root cause of the source of the problem

• Student Drop out prediction based on region / family status / locality
• Employment Generation prediction across industries
• Crime Analysis and detection based on income /segment/education
• Income Tax Evasion across  income / segment / education / region
• Predict potential offenders based on financial activities/network connections

Happy Mastering DL!!!