Performance of Windows10

• Optimize disable hardware acceleration in browser/teams
• Disable unused services RTX voice, Use only when environment noisy
• Keep updated drivers 
• Keep the Laptop well ventilated
• Avoid extensions if they heat up

We pay so much money, Running browser, teams, ppt, and outlook system sounds like a fire engine :( crazy software, Reflects there is no sync up for hardware and software. Make customers buy new laptops every few months.

Keep Checking!!!

Men's Hair Styles Approach

Paper #1 - Real-time deep hair matting on mobile devices

Key Notes

• Hand-crafted features for segmentation.
• Employ simple pixel-wise color models to classify hair.
• Fully Convolutional MobileNet Architecture for Hair Segmentation

• HairSegNet
• Pre-trained weights on ImageNet, we dilate all the kernels for the layers with updated resolution by their scale factor
• Upsampling is performed by a simplified version of an inverted MobileNet architecture
• a loss function that promotes perceptually accurate matting output

• HairMatteNet runs twice as fast compared to HairSegNet

Paper #2 - Intuitive, Interactive Beard and Hair Synthesis with Generative Models

Key notes

• Edge detection or image gradients would be an intuitive approach
• Generative adversarial networks (GANs)
• Two-stage pipeline
• First stage focuses on synthesizing realistic facial hair
• Texture synthesis techniques
• pixel-based methods
• stitching-based methods

Generative adversarial networks (GANs) [26] has inspired a large body of high-quality image synthesis and editing approaches

Two Stage Network

• The first stage synthesizes the hair in this region.
• The second stage refines and composites the synthesized hair into the input image.

Close-up images of high-resolution complex structures fail to capture all the complexity of the hair structure, limiting the plausibility of the synthesized images

StyleGAN

Paper - Link

• Generator embeds the input latent code into an intermediate latent space

Deep Convolutional Generative Adversarial Network 

• During training, the generator progressively becomes better at creating images that look real, while the discriminator becomes better at telling them apart. 
• The process reaches equilibrium when the discriminator can no longer distinguish real images from fakes.
• Both the generator and discriminator are defined using the Keras Sequential API.

How to code a Generative Adversarial Network (GAN) in Python

Paper #3 - PROGRESSIVE GROWING OF GANS FOR IMPROVED QUALITY, STABILITY, AND VARIATION

• Our primary contribution is a training methodology for GANs where we start with low-resolution images, and then progressively increase the resolution by adding layers to the networks
• When training the discriminator, we feed in real images that are downscaled to match the current resolution of the network

• MULTI-SCALE STATISTICAL SIMILARITY FOR ASSESSING GAN RESULTS
• Intuitively a small Wasserstein distance indicates that the distribution of the patches is similar

Paper #4 - Training Language GANs from Scratch

ScratchGAN code 

• Generator architecture and reward structure
• Large Batch Sizes for Variance Reduction
• Unlike image GANs, ScratchGAN learns an explicit model of data

Paper #5 - A review of Generative Adversarial Networks (GANs) and its applications in a wide variety of disciplines - From Medical to Remote Sensing

• Generative Adversarial Network (GAN) colab
• Conditional_gan colab
• InterFaceGAN - Interpreting the Latent Space of GANs for Semantic Face Editing
• StyleGAN — Official TensorFlow Implementation
• Hairstyle Transfer — Semantic Editing GAN Latent Code

CelebHair: A New Large-Scale Dataset for Hairstyle Recommendation based on CelebA

More Reads

• Hair Color Digitization through Imaging and Deep Inverse Graphics
• Intuitive, Interactive Beard and Hair Synthesis with Generative Models
• Progressive Color Transfer with Dense Semantic Correspondences
• Local Facial Attribute Transfer through Inpainting
• Generative Single Image Reflection Separation
• ReflectNet - A Generative Adversarial Method for Single Image Reflection Suppression
• A Style-Based Generator Architecture for Generative Adversarial Networks
• K-HAIRSTYLE: A LARGE-SCALE KOREAN HAIRSTYLE DATASET FOR VIRTUAL HAIR EDITING AND HAIRSTYLE CLASSIFICATION
• AttGAN: Facial Attribute Editing by Only Changing What You Want
• HairCLIP: Design Your Hair by Text and Reference Image
• fAshIon after fashion: A Report of AI in Fashion
• Keypoints-Based 2D Virtual Try-on Network System
• StyleCLIP: Text-Driven Manipulation of StyleGAN Imagery
• VOGUE: Try-On by StyleGAN Interpolation Optimization
• Dressing in Order: Recurrent Person Image Generation for Pose Transfer, Virtual Try-on and Outfit Editing

Keep Thinking!!!

What does it take to build a good enough CNN model ?

This question came up during class. There is no direct answer. It depends on domain / data / use case / number of classes. Guidelines of recommendations are

• Number of Layers of convolution, Experiment with VGG16 / 19 to get started
• Balacing imbalanced datasets
• Depth of Layers - 32/ 64/ 256 as needed
• Relu, Experiment, Customize on activation functions
• Adjusting learning rates / Loss functions
• Early Stopping / Dropouts for regularization
• Domain Relevant Augmentation
• Transfer Learning Approach

Keep Thinking!!!

Ideation - Techniques

What does it take to build a production grade working product ? 

Clarity of business impact of idea

Competitive products in the market to draw the source of market/landscape/potential

Research papers to understand basics of implementation. Collect all possible ideas / Strategies

Take all the tools of the idea, Have the clarity to translate vision to working tech with steps/models/data/approach, Bridging between great vision vs transparent planning and predictable execution

Consistency in vision, execution, innovating every problem with your experience, and open to accept failures and relearn as needed. You will build the learning path that best works for you. Skill/consistency is long-term. It goes beyond copy-paste/certifications.

Keep Thinking!!!

CDP - PDP - DDP - RDP - FDP - HDP

All the data in silos vs having something relevant, anonymized, and useful like a knowledge repository (Not for Profit)

• CustomerDataPlatform  
• DiseasesDataPlatform 
• PatientDataPlatform 
• FinancialDataPlatform
• EducationDataPlatform
• MentalHealthDataPlatform

Helping each other in sharing their experiences, Not for profit but for benefit of humanity.

Bridging the 2/3rd world while the 1/3rd goes ahead with Adoption, AI, and Better lifestyles.

Keep Thinking

Building products

Every product that we use in AI / ML work seems to be built by a core set of a small number of focused Teams.

• Streamlit - 21 Employees
• BentoML - 8 Employees
• DagsHub - 13 Employees
• MLFlow - < 10 Employees

We need a focused team of passionate people. Employee count is just a number. Focus, Focus, Focus, Consistency, and Persistence. 

Keep Thinking!!!

Decode Content Engagement across apps / Alive and productive or Alive and Unhappy

Across different platforms / Apps, How the traps are set for Content engagement and amplification. How do you draw the line between #healthy consumption vs #dangerous consumption?

Youtube

• The subscribe option that shows up regular daily videos
• Updates to your favorites - Paranormal, travel vlogs, stocks, etc..
• This easily takes up time before sleep (Adds to your menu list with constant follow up / curiosity)
• Walking time is almost gone
• Recommendations of video list to keep you scrolling

Swiggy

• The notification / offers trap
• Beautiful recipe pictures
• Add more to the menu and get a 30% discount with X value order
• Always this comes in mind, Even if you don't cook you can swiggy it anytime food delivered 
• 50 rs if I walk for 5 mins, 300 rs for ordering from smartphone :), One-click at cost :)

Twitter

• Search by trending tags and keep hearing opinions
• Look for opinions similar to your beliefs
• After a TV show lookup for trending topics to engage more in content

TV

• Debate shows / Talk shows / comedy shows / Expedition X
• People know the type of content/type of engagement
• Although unverified but content that excites with uncertainty
• Breaking news of constant repetition and sensation

Work (Some useful work)

• Read papers / Experiment code
• Look for solutions/references
• 30% is useful, 70% is a useless Internet content

Whatsapp / Telegram

• Groups / messages / discussions
• Status in WhatsApp to know things around us
• Constant forwards / motivation / count of unread messages / groups
• Marketing forwards / calls from marketing/loans offers / Numbers that get shared with data registered in the apps

"Many time blocks of the day were used only for work before social media now take up smartphones". Everything consumes time. 

What we forget

• Reminder I waste time and I age losing my precious time and thoughts, How slowly we change our routine based on the content we consume, gradual increase in content consumption with the personalized recommendations
• A reminder this content is for attention and not really true always
• The food in the picture is good for taste but bad for the heart
• Everybody cannot afford to travel, focus on work
• Don't live or long for Life that you don't have
• Internet is a garbage yard, the more you travel without a plan the more you end up with trash
• Eye problems, sedentary lifestyle everything will be an issue
• Swiggy food looks good in pictures, and discounts but in quality often consuming it what about cholesterol?
• Die early, get more health issues, feel emotionally low
• Every age group kids/teens/adults / aged everyone has similar to their preferences and choices of apps for engagement, The changes come in all segments. How much we use for learning vs entertainment? How much do we get often distracted?
• The small amounts of loss of focus how impact cascades into habits and loss of productivity

We have slow poison pills everywhere around us. Alive and feeling unhappy vs #alive and #focused vs alive and #purposeful. We need to set boundaries in our usage/consumption and be consciously aware of purpose/usage, or else we will end up as busy and unproductive.

Good read - Link

• Fake news. Misinformation. Mental health crises. Election meddling. Extremism. Partisanship. Eating Disorders. Body Dysmorphia. Vulnerability porn.
• Parasocial relationships4 where a single person can make thousands (or more) feel as though they are interacting on a personal, one-on-one level.
• #Socialmedia companies hire #programmers and #datascientists to operate “engaging user interfaces” that strip mine your attention.
• "...you’re not the customer, you’re the product."

Keep Thinking!!!

Read - Write - Experiment - Unlearn - Relearn

Over a point in time, once we collect/analyze trends/patterns we will have our perceptions of how things are shaping up

• Customer data patterns
• Product adoption vs predictions
• Gartner hype vs Reality

Bring it down to 300 feet when solving day to day vision, forecasting, recommendations, and data problems

• Alternatives
• Algos / Data challenges
• Vision options models / custom training / alternative approaches

The mind always keeps things looking out 

• Running thoughts of options / references / ideas
• Experimenting with codes and evaluating pros/cons
• Planning for next steps / next aspects of product building
• Agile / Calculated risks / /Proactive to spot failures
• Plan for the team, Don't miss anything to waste team efforts
• Learn and upskill self, to guide / proactively spot red flags
• Use experience to apply and identify risks
• Being productive not just busy needs a running set of ideas/thoughts
• Reduce the gaps between Great planning vs Poor Execution

The mind always keeps on what next to anticipate, and prepare for .....

Keep Going!!!

My observations from - How different companies ship code to production

Microsoft days 

• Way back 14 years ago
• Automated build
• CI / CD
• DEV - Test - Support - Prod
• Environment for each function
• Bugs / Triage 
• Only Stable code will goto prod :)
• Bug bashing, Support bashing...
• Comes next to Amazon but has tighter QA
• Phased / frequent releases
• Expensive rollbacks so had tighter QA

Amazon days

• Saas world / 10 years ago itself
• Everything as service
• One click deployment
• Ton of A/B testing/experiments
• Weblabs / weblabs / Test ton of features same time
• The best of several teams working together
• One-click rollback too so was easier to revert back
• Weekly deployments

Sensormatic

• Like windows service packs 
• Once every 6 months
• Pure offline mode installation
• Hardware + Software combo
• Customization / Customer specific feature
• Bit far compared to Amazon / Microsoft practices

Consulting

• CI / CD
• Up to date with the latest tools in the market :)
• Containerized / dockerized / deployed in cloud
• Fast / Fast / Fast. 6 to 8 weeks
• Closer to startup type culture :)

The quality bar is relative to maturity of product, number of active customers, Complexity of implementation, expectations from beta product vs production system

More reads - Link

Keep Thinking!!!

Learning from code reviews

An interesting read from link

My favorite list

Perspective #1 - You don’t need hundreds of engineers to build a great product

[Siva] - Have a set of ideas, experiment, fail, learn, unlearn, relearn, and Build a vision of the product not under the pressure of timelines

Perspective #2 - Simple Outperformed Smart

[Siva] - Start to crawl before you learn to run

Perspective #3 - Our highest impact findings would always come within the first and last few hours of the audit.

[Siva] - Functionality, Scalability, Performance matters

Perspective #4 - Business logic flaws were rare, but when we found one they tended to be epically bad

[Siva] - Product is a for customer need not for experimenting technology. Build what is needed for the customer, provide the customer experience

Perspective #5 - Quick turnarounds on fixing vulnerabilities are usually correlated with general engineering operational excellence.

[Siva] - Quality ideas come from the domain, data, and functional understanding. Think from the long term no near term fixes

Good Read - Link

• How does the number of reported defects in source code files correlate to source code quality?
• How much longer development time is needed to resolve an issue in files with low-quality source code?
• To what extent is the code quality of a file related to the predictability of resolving issues on time?

Keep Thinking!!!