Code execution should take place on Google Colab in the .ipynb format.
Phase I of the coding contest begins on 17th August 2023 and concludes on 25th August 2023.
On 20th August 2023, participants will receive a Google Drive folder link to submit their codes.
On 20th August 2023 Test datasets will be provided in your Google Drive folder link
The selection period will run from 26th to 30th August.
The announcement of the Phase II selected contestants will be made on 31st August 2023.
Develop an algorithm for Pedestrian Attribute Recognition that can accurately distinguish among pedestrians from various age groups: child, adult, and senior. The algorithm should also segment and classify their clothing and accessories as visual cues. Design a solution that achieves a minimum accuracy of 85% on this dataset. Additionally, elucidate how your model addresses prevalent challenges such as occlusions, variations in clothing styles, and different camera angles.
Construct an algorithm capable of classifying and predicting species from a plant dataset. The objective is to utilize artificial intelligence to accurately categorize plant images into their taxonomic groups while forecasting their specific species. Submissions should report the F1 score, precision, and recall.
Your task is to:
Preprocess the dataset, including resizing images to a consistent size and splitting it into training, validation, and test sets.
Design and implement a deep neural network architecture suitable for this multi-class classification problem.
Address challenges such as imbalanced classes, variations in lighting and background, and potential occlusions.
Train your model using the training set, optimizing the chosen performance metric.
Evaluate your model on the validation set and fine-tune hyperparameters as needed.
Provide an analysis of your model's performance, discussing its strengths and limitations.
You can utilize publicly available datasets. Highly recommended datasets links that can be utilized are provided below.
Note: If using any other datasets, mention the datasets used and link for downloading the datasets
Kindly present your submission with well-documented code, elucidating your rationale behind design choices, encountered hurdles, and innovative strategies employed to bolster model efficacy amidst the nuanced botanical image landscape.