Use examples when they help make things clearer.

Land classification using deep learning CNN (Convolutional Neural Networks) involves analyzing satellite imagery and identifying patterns that indicate different types of land use, such as forests, agricultural fields or urban areas. By training a neural network on labeled data, researchers can create models that accurately classify new images based on these patterns. This information is useful for environmental monitoring and resource management.
For example, the “Land Classification Using Deep Learning CNN” project uses satellite imagery to identify areas in Brazil that are most suitable for agriculture. By analyzing factors such as soil type, rainfall patterns, and vegetation cover, researchers can create models that accurately predict which regions will be most productive for crops. This information helps farmers make informed decisions about where to invest their resources and how best to manage their land.
Another example is the “Satellite-Led Liverpool” project, which uses remote sensing data to measure living environment deprivation in urban areas. By combining satellite imagery with machine learning algorithms, researchers can identify areas that are most in need of investment and resources. This information helps policymakers make informed decisions about how best to allocate funds for economic development programs.
Weakly & semi-supervised learning are two methods of machine learning that use both labeled and unlabeled data for training. Weakly supervised learning uses weakly labeled data, which may be incomplete or inaccurate, while semi-supervised learning uses both labeled and unlabeled data. Weakly supervised learning is typically used in situations where labeled data is scarce and unlabeled data is abundant. Semi-supervised learning is typically used in situations where labeled data is abundant but also contains some noise or errors. Both techniques can be used to improve the accuracy of machine learning models by making use of additional data sources.

For example, “MARE” uses weakly supervised multi-attention resunet for semantic segmentation in remote sensing. This technique allows researchers to train a model on unlabeled satellite imagery and then fine-tune it using labeled data. The result is a more accurate and efficient model that can be used to classify land use patterns with high precision.
Another project, “WHU-Stereo,” is a large-scale dataset for stereo matching of high-resolution satellite imagery that includes several deep learning methods for stereo matching such as StereoNet, Pyramid Stereo Matching Network & HMSM-Net. These techniques allow for more accurate and efficient stereo matching without requiring manual labeling or segmentation of data.
Finally, “Photogrammetry-Guide” is a guide covering photogrammetry that includes applications, libraries, and tools to make you a better photogrammetrist. This resource can be useful for those interested in learning more about the field and improving their skills.