IEEE-CSE Projects 2018

Aim:   To build a dataset to learn a powerful CNN-based SICE enhancer from multi-exposure images. Proposed system: propose to use the convolutional neural network (CNN) to train a single image contrast enhancement (SICE) enhancer. A CNN can be easily trained as the SICE enhancer to improve the contrast of an under-/over-exposure image. Overcome: One key issue is how to construct a training data set of low-contrast and high-contrast image pairs for end-to-end CNN learning      Advantage:     CNN based SICE method learns a complex nonlinear mapping function to map a low-contrast (either underexposure or over-exposure) region to a good contrast region.     powerful CNN-based SICE enhancer, which is capable of adaptively generating high quality enhancement result for a single over-exposed or underexposed input image.        Existing System:   Previous SICE methods adjust the tone curve to correct the contrast of an input image. often fail in revealing image details

Aim: To provide an answer for the question: what are the best values for these two properties (spatial resolution and color) of digital cameras to use in vision-based vehicle classification systems? In other words, how many pixels are sufficient to build an accurate vision-based vehicle classification system? Aim: To study the effect of spatial resolution on the accuracy and performance of image classification methods. Proposed System: present a comprehensive study of the effect of these two spatial characteristics ( Dimension and color ) of digital images on the vision-based vehicle classification process in terms of accuracy and performance. For Bag-of-Visual Words (BoVW) [13], Vector of Locally Aggregated Descriptors (VLAD) [14] and Fisher Vector (FV) [15] it is recommended to use the DSIFT descriptor. The ResNet/Convolutional Neural Network (CNN), architecture is considered the top classification method with an accuracy rate of 95.12% compare to BoVW, VLAD an