#LSPPDay52@lftechnology
Today I got hands on project experience on complete computer vision where I evaluated traditional vs. deep learning baselines, generative modeling and data augmentation and modern vision transformers.
#60daysoflearning2026#LearningWithLeapfrog
#LSPPDay51@lftechnology
Today I learned how model can learn shortcuts from the similar background of training samples neglecting the actual shapes and features. Also learned the difference of kernels of ResNet and conv from scratch.
#LearningWithLeapfrog#60daysoflearning2026
#LSPPDay49
Today I revised convolution, its distinct feature, hierarchy of features, why ReLU is non linear activation function. I also learned about ouput size of CNN, transfer learning, padding and stride property.
@lftechnology#60DaysOfLearning2026#LearningWithLeapfrog
#LSPPDay48
I continued my major project, there were some merging conflict in the repo, so I solved it. Changed my branch name and worked in the automated report generation for each computation of the stations.
#60DaysOfLearning2026@lftechnology#LearningWithLeapfrog
#LSPPDay47
Today I worked on my project of wind aware DGNN for AQI, where I normalized the stations and mapped the names to index, then I calculated the distance matrix of all 48 stations using Haversine formula.
@lftechnology#60DaysOfLearning2026#LearningWithLeapfrog
#LSPPDay44
Today I just learned the mathematics behind the complex backpropagation. Learned the crisp difference of forward and backward propagation. Did the actual calculations in this backpropagation with XOR gate.
@lftechnology#60DaysOfLearning2026#LearningWithLeapfrog
#LSPPDay42@lftechnology#60DaysOfLearning2026#LearningWithLeapfrog
Today I got clarity on why we need to convert the color space when working with OpenCV and matplotlib, the mandatory dimension of an image and why Hue value in OpenCV is half of the standard HSV hue.
#LSPPDay39@lftechnology#60DaysOfLearning2026#LearningWithLeapfrog
Today I was doing hands on coding of building CNN layers, optimizing it using GridSearch and optuna. I learned about internal covariate shifts and activation mangnitude stability provided by normalization.
#LSPPDay37#60DaysOfLearning2026#LearningWithLeapfrog
Today I read the same research paper DGNN-AEA further and learned about decomposition of adjacency matrix into weight and topology matrix, about multi graph stitching.
#LSPPDay36#60DaysOfLearning2026#LearningWithLeapfrog
Today I learned about sigmoid, its mechanism of backpropagation and vanishing gradients. Learned the logic to build the NN from scratch using __init__ and forward() function.