Classification of weather situations on single color images
read more
Citations
Auxiliary Tasks in Multi-task Learning.
Two-Class Weather Classification
Weather classification with deep convolutional neural networks
Automated driving recognition technologies for adverse weather conditions
A CNN–RNN architecture for multi-label weather recognition
References
Histograms of oriented gradients for human detection
Object recognition from local scale-invariant features
Beyond Bags of Features: Spatial Pyramid Matching for Recognizing Natural Scene Categories
Video Google: a text retrieval approach to object matching in videos
Related Papers (5)
Frequently Asked Questions (11)
Q2. What are the future works in "Classification of weather situations on single color images" ?
Future work will expand C by adding other weather situations like fog to their database. Improvements of the overall classification results could be achieved by further in-depth studies to non linear SVM-kernels.
Q3. How does the approach achieve low error rates?
The approach achieves low error rates of less than 1% for the distinction between clear weather and heavy rain and even acceptable error rates for the three-class-case.
Q4. What is the principle of a linear SVM?
In principle, a linear SVM generates a hyperplane in the descriptor space D and classifies descriptors by calculating on which side of the hyperplane the descriptor vector (=point) lies.
Q5. What can be done to improve existing algorithms?
Specialized methods on certain weather situations can then be invoked based on the classification result to improve existing vision algorithms.
Q6. How many features are extracted from the HSV?
the authors combine all histograms into one extended descriptor vector, so the authors get vector v = (v1, ..., vn) with n = (13 ROIs) ∗ (5 features) ∗ (10 bins) = 650 scalar elements describing the image.
Q7. What is the significance of each dimension of descriptor space D?
As mentioned in section III, parameter w of the SVM tells us the significance of each dimension of descriptor space D. In their experiments, all feature weights are evenly distributed, that means not one feature alone or any combination of some features is able to achive high discrimination, the descriptiveness lies in the combination of all proposed features.
Q8. What features are evaluated within each ROI?
Within each ROI, several features are evaluated: local contrast, minimum brightness, sharpness, hue and saturation, detailed below.
Q9. What is the effect of the weather on the pixel?
Equation (1) implies that the irradiance and thus the brightness observed by each pixel of the sensor is altered by two fundamental scattering phenomena: attenuation and scattered light.
Q10. What is the effect of the weather on the local contrast?
Inserting (1) in (3) yieldsC = ρmax − ρminρmax + ρmin + 2(eβd − 1) . (4)As a result, the local contrast solely depends on scene point properties (which remain constant), distance d and the scattering coefficient β.
Q11. How did the authors investigate the results in Table I?
The authors investigate the results in Table The authorin more detail by applying binary classification to the image sets, that means the authors only take images from 2 classes.