course 22 P1 - Pass 2

[manisnesan]

• Object Detection Lesson 8 P2 2018
• Lesson 3
• Lesson 2
• Lesson 1
• Lesson 0

Target Start : Mar 29
Daily 30-1 hr Continuously until I Finish it.

Focus Areas

• Project : Humpback Whale Identification
• Data Gathering and Validation

Strategies

• Learn: Why should you read lecture summaries and questionnaires before/after studying the lecture?
• Practice : Why a real world DL practitioner spend most of the valuable/productive time preparing data rather than tweaking models?

Resources for effective learning

• Five techniques - YouTube - priming (outline),at event (ask Why and How Qs), post event revision within 12 hrs, techniques for review(non verbal, mindmap) preexam revision (teaching, tests recall)

[manisnesan]

Apr 6, 7

• Lesson 3. Neural net foundations

Apr 5

• Explored Icevision getting started but faced a blocker due to cuda version not compatible with the library in order to experiment.

Apr 4

• Explored walkwithfastai Lesson 6 Object Detection notebook and the corresponding lesson from 2018 course for 30m. I haven't explored Object Detection in fastai. Lots of new areas to explore especially Data Preparation using annotations, Model Types, Metrics etc. I may need to revisit this once I finish Pass 2.

Apr 3

• Lesson 1. JS based Classifier applications are simple & easy way to showcase the work to users for MVP needs.

Apr 2

• Code: Took a detour by learning about Object Detection using https://walkwithfastai.com/Object_Detection with the help of PASCAL dataset. Colab Ran

Apr 1

• Rewatched Lesson 1. One aha moment: Train the model first before cleaning the data using the model's learning capability. This is now true in the LLM model space whether the model is trained by us or someone else. Role of the model is to generate quality data.

Mar 31

• Rewatched Lesson. Reviewed Summary. Whale Identification Kaggle Challenge would be nice candidate project to tackle as part of deliberate practice in this pass.

Mar 30

• Replicated the experiment and reproduced the results using a different dataset for 'is it a bird'. See Colab

Mar 29

• Reviewed lesson 0 again. Key takeaway is "Finish it"
• I have done as part of P1. Completed Watch the lecture, Running the code and experiment.
• I will be focusing on Reproducing the results and Repeat with different dataset.

[manisnesan]

Tricks

Control how the output is displayed in python libraries. Learn more about them here

np.set_printoptions
torch.set_printoptions
pandas.set_printoptions