2020 International Conference on Innovative Trends in Information Technology (ICITIIT) 978-1-7281-4210-4/20/$31.00 ©2020 IEEE Abstract— Image-based object retrieval has numerous applications in the field of machine vision to inquire from an appropriate image or video sequence for a given query object. The object retrieval task is conventionally carried out by a set of handcrafted algorithms, which provides image depictions in the fashion of visual characteristics. During the last decade, an extensive change has been practiced to describe visual content from handcrafted characteristics to the application of machine learning approaches and to the real layout of the image descriptors. The extensive movement is based on Convolutional Neural Networks (CNN) which is popularly known as Deep Learning. This proposed work deals with a combination of both conventional and machine learning approaches to retrieve an image object from a given dataset. This is done by a series of activities such as feature extraction and storage of training images, query image selection and feature extraction. Similarity matching between database and query image features. Final retrieval is based on the objectness score. Index Terms— Image Object Retrieval, SIFT, R-CNN, Feature Extraction, Feature Matching. I. INTRODUCTION Retrieval of an object from a given scene is different from traditional content-based image retrieval (CBIR). Retrieval of image objects rather than image targets on fetching object-level details of a given image, which generally present in various ambiances. The ultimate goal of image object recovery is to return specific queried object-level content from an image repository. Efficient retrieval of image objects is treated as much difficult when it deals with a large volume of data because destination commonly occupies meager parts on images. If an object-based image is comparatively tiny in size and the picture is crowded with many other objects, then the retrieval process became tedious. The objectness measure acts as a class-generic object detector. Retrieval of relatively small inquired image object from a given picture is one of the difficult tasks in image object retrieval. The goal of the current study is to develop an efficient object-level retrieval model to address the background intervention problem, the region proposal generation, classification re-ranking and finally object retrieval based on objectness score. Object retrieval plays a major role in the present era of artificial intelligence and machine vision. Object retrieval has various applications in our day-to-day life. Computer vision is commonly used in multiple sectors because of the improvements in deep learning methods. Computer vision is a multidisciplinary track that manages with how machines can acquire high-level knowledge from digital images. Machine vision is made possible through a series of tasks. The outcome of the computer vision task is the distillation of relevant information, to produce some meaningful decisions. The object retrieval task is conventionally carried out by a set of handcrafted algorithms, which provides image depictions in the form of visual features. Image feature representations made by Scale Invariant Feature Transform (SIFT) or Speeded Up Robust Features (SURF) have served as essential in computer vision in the time of first decade of 2000s. Even though, for the past few years, a massive development has been accomplished to describe visual values from handmade characteristics to the application of machine learning approaches and the original design of the image descriptors. The massive trend is depend on Convolutional Neural Networks (CNN) which is popularly known as Deep Learning. Computer vision provides specific applications in the area of object retrieval from a given scene. Object retrieval task generally carried out by two steps, searching for image objects and its retrieval [2, 3]. The main objective of this paper is accomplished by four different steps. Firstly, generate an effective object proposal using a bounding box. This will make the object retrieval task easy. Secondly, classify the detected objects in the context of the query image. Image classification is used to categorize the given images into different classes. Thirdly perform an object-level re-ranking based on the objectness score in the given dataset. Finally, retrieve the image object from a large collection of image data. Every object appeared on an image scene has its object boundary, which is termed as region proposals. As an initial stage we aim to identify various object region proposals in a given image. There are many distinct methods available to generate region proposals. The basic idea of generating region proposals is based on bounding boxes. Each bounding box represents segmented portions from a given image. The feature should be extracted from the images only when the proposal selection is been done. It will maintain the variable-sized image input into fixed-sized visual values. The feature extraction process is done by any of the conventional methods such as filter-oriented, histogram-based or deep learning-based models [3, 4, 5 and 6]. The next section will give a detailed idea on different object retrieval schemes both in conventional and machine learning approaches. Object Retrieval in Images using SIFT and R-CNN Amitha I C N K Narayanan Department of Information Technology Indian Institute of Information Technology Kottayam Kannur University Kerala India. Valavoor (P.O.) Kottayam-686635 Kerala. amithaic@sngcet.org nknarayanan@iiitkottayam.ac.in