In today's world, image processing plays a crucial role across various fields, from scientific research to industrial applications. But one particularly exciting application is image captioning. The potential impact of effective image captioning is vast. It can significantly boost the accuracy of search engines, making it easier to find relevant information. Moreover, it can greatly enhance accessibility for visually impaired individuals, providing them with a more immersive experience of digital content. However, despite its promise, image captioning presents several challenges. One major hurdle is extracting meaningful visual information from images and transforming it into coherent language. This requires bridging the gap between the visual and linguistic domains, a task that demands sophisticated algorithms and models. Our project is focused on addressing these challenges by developing an automatic image captioning architecture that combines the strengths of convolutional neural networks (CNNs) and encoder-decoder models. The CNN model is used to extract the visual features from images, and later, with the help of the encoder-decoder framework, captions are generated. We also did a performance comparison where we delved into the realm of pre-trained CNN models, experimenting with multiple architectures to understand their performance variations. In our quest for optimization, we also explored the integration of frequency regularization techniques to compress the "AlexNet" and "EfficientNetB0" model. We aimed to see if this compressed model could maintain its effectiveness in generating image captions while being more resource-efficient.

Our immune system is a complex network that consists of cells, tissues, and organs that operates concurrently to shield our body from millions of diseases, causing bacteria, parasites, and viruses. For the identification of different kinds of hematological disorders, the accurate identification of various White blood cells (WBC) is necessary for classification purposes. Most of the diseases can be diagnosed by the numbers and sizes of white blood cells found in a blood smear. A drastic change in a particular WBC count relative to the standard range provides us a hint about being attacked by distinct enzyme. As the incorrect segmentation of cells leads to inaccurate disease detection, it demands utmost significance that this process is performed in the best effective way. Still now, in many medical centers the detection and categorization of WBCs is performed manually by experts. As there remains a great probability of error due to manual classification, automatic systems should be designed in such a way that there will be a very minimal error rate as compared to the manual. With this aim, in this paper, a renowned methodology named Deep learning is proposed to conduct the whole classification process automatically applying an improved lightweight convolutional neural network which has been implemented for both multiclass and binary classification with an accuracy rate of 98.63% and 91.95% respectively.

White blood cell (WBC) count in our bloodstream plays a notable role in the diagnosis or prognosis of various blood diseases like acute lymphoblastic leukemia, heart diseases, or infections. While a radical change in white blood cell count comparative to the baseline provides us a sign that our body is being attacked by an antigen, it protects us from various infectious diseases as well. A particular type of white blood cell count tells us about being affected by a specific antigen. In medical diagnosis, it is of extreme importance to distinguish the different white blood cell components efficiently. An incorrect diagnosis may also become the cause of death. At present, in most of the medical centers, this classification is done manually by experts, which is time-consuming. Though some semi-automated procedures are being proposed where feature extraction is done manually before classifying automatically using microscopic blood smear images, it is still time-consuming and tedious. In recent years ANN, CNN-RNN, and also fused CNN models have been employed to perform the WBC classification. For making the procedure more effective, we propose a Deep Learning methodology to perform the classification applying a convolutional neural network model with a higher accuracy rate and a lower number of parameters compared to the state-of-the-art approaches.