Computer Vision Lab Experiments
Experiments
38 results
- Converting the image to greyscale and reading the histogramGrayscale and histogramConvert a colour image to greyscale and compare the histogram before and after.Image processingBeginner
- Brightness and contrastTry the effect of α and β on a greyscale image and watch for saturation.Image processingBeginner
- Revealing shadows with gamma correctionUse γ below 1 to bring out dark details.Image processingBeginner
- Gaussian versus medianGaussian vs median filterCompare linear and non-linear smoothing at the edges.Image processingBeginner
- The bilateral filter preserves edgesEdge-preserving bilateral filterSmooth the image without blurring the boundaries.Image processingBeginner
- Sobel directionsSee how the horizontal Sobel captures vertical edges and vice versa.Image processingBeginner
- The two Canny thresholdsAdjust the two thresholds and watch the edge connectivity.Image processingBeginner
- Otsu on the coinsSeparating the coins from the background with an automatic threshold.Image processingBeginner
- Cleaning the mask with morphological operationsCleaning a mask with morphologyA chain: Otsu threshold, then closing.Image processingBeginner
- CLAHE on the retinal imageImprove the local contrast to bring out the vessels.Image processingBeginner
- Aliasing when downscalingCompare interpolation methods under heavy downscaling.Image processingBeginner
- Convolution by hand, step by stepFollow each cell of the feature map.Convolution and neural networksBeginner
- The effect of stride and poolingHow the dimensions change with stride and pooling.Convolution and neural networksBeginner
- Classify the coffee cupRun a real model and read the top five classes.Image classificationBeginner
- The confidence threshold in detectionHow many boxes remain at each threshold?Object detectionBeginner
- Semantic segmentation and mask overlayColour every pixel by its class and move the opacity.Image segmentationBeginner
- Grad-CAM for different classesDoes the map change when you explain the second class instead of the first?Visual explanation and attention mapsAdvanced
- Grad-CAM++ versus Grad-CAMCompare the two methods on an image with repeated objects.Visual explanation and attention mapsAdvanced
- The effect of the occlusion patch sizeThe trade-off between spatial resolution and time.Visual explanation and attention mapsAdvanced
- The stability of the LIME explanationHow does LIME change with the number of samples?Visual explanation and attention mapsAdvanced
- Attention Rollout on DeiTAn attention flow map in a real vision transformer (needs the DeiT pack).Visual explanation and attention mapsAdvanced
- Agreement between explanation methodsMeasure the correlation coefficient between Grad-CAM, Occlusion and LIME.Visual explanation and attention mapsAdvanced
- Learned feature mapsCompare hand-designed maps with maps learned by MobileNetV2.Convolution and neural networksAdvanced
- The effect of the NMS thresholdHow do overlapping boxes change with the IoU threshold?Object detectionAdvanced
