Bracelet ✓
Book
Pen
Pumpkin
Microphone ✓
Dice
Water Bottle ✓
Sunglasses ✓
Scissor
Through my experiment, I realized the complexities and nuances involved in machine learning and image classification. In the experiment I conducted using the ml5.js script and MobileNet model to classify various objects in my room, I noticed that the model sometimes returned multiple results for a single object, with varying degrees of confidence scores. While it accurately recognized 5 out of the 10 objects I tested, I found it intriguing that it sometimes suggested potential results for the others. I feel like the reason why causes the multiplicity of answers could be attributed to the objects having characteristics that are common. There are definitely limitations to the dataset; its ability to correctly classify objects might be limited to the diversity and quality of data it was trained on.
Throughout my exploring of the websites, I have found something in common. First of all, they can be used to collect large volumes of data from users. For example, a website could gather user-generated content, such as text, images, or even audio, and use it to create or enhance machine learning datasets. They can also preprocess data before feeding it into machine learning models. They also aggregate data from multiple sources, curate datasets, and perform data augmentations. Some of the data sets are authorized by authority which further enhance the authenticity and reliability. I found the COVID-19 Open Research Dataset Challenge and Chest X-Ray Images (Pneumonia) from kaggle to be very significant and valuable resource. The X-ray dataset can be used to train and develop machine learning models for pneumonia detection from chest X-ray images. This can assist healthcare professionals in automating the initial screening process. Researchers can use CORD-19 to explore critical scientific questions related to COVID-19, as outlined by authoritative bodies like the National Academies of Sciences, Engineering, and Medicine and the World Health Organization. However, in considerate of of ethical issues, researchers and organizations should commit to using the data responsibly. This includes ensuring that any AI or NLP applications developed do not harm individuals, communities, or society at large. Moreover, researchers should be transparent about their methods, findings, and any potential conflicts of interest since accountability is important to maintain the integrity of research conducted using this dataset.
Comments