ABSTRACT:

The key challenge for few-shot semantic segmentation (FSS) is how to tailor a desirable interaction among support and query features and/or their prototypes, under the episodic training scenario. Most existing FSS methods implement such support/query interactions by solely leveraging plain operations – e.g., cosine similarity and feature concatenation – for segmenting the query objects. However, these interaction approaches usually cannot well capture the intrinsic object details in the query images that are widely encountered in FSS, e.g., if the query object to be segmented has holes and slots, inaccurate segmentation almost always happens. To this end, we propose a dynamic prototype convolution network (DPCN) to fully capture the aforementioned intrinsic details for accurate FSS. Specifically, in DPCN, a dynamic convolution module (DCM) is firstly proposed to generate dynamic kernels from support foreground, then information interaction is achieved by convolution operations over query features using these kernels. Moreover, we equip DPCN with a support activation module (SAM) and a feature filtering module (FFM) to generate pseudo mask and filter out background information for the query images, respectively. SAM and FFM together can mine enriched context information from the query features. Our DPCN is also flexible and efficient under the k-shot FSS setting. Extensive experiments on PASCAL-5 i and COCO20i show that DPCN yields superior performances under both 1-shot and 5-shot settings.

EXISTING SYSTEM :

Semantic segmentation is a classical computer vision task which aims to give pixel-wise prediction for an input image. Recently, various networks [14] have been actively designed to further improve the semantic segmentation results. For capturing more contextual information, dilated convolution [33], pyramid pooling [40], and deformable convolution [3], are proposed to enlarge the receptive filed. Meanwhile, some models leverage attention mechanisms [6, 27, 34, 37] to capture long-distance dependencies for semantic segmentation, which reach state-of the-art performances. However, these semantic segmentation approaches still fail to preserve their initial performances when insufficient training data is provided

EXISTING SYSTEM DISADVANTAGES:.

1.LESS ACCURACY

2.LOW EFFICIENCY

PROPOSED SYSTEM :

To address the above challenges, we propose a dynamic prototype convolution network (DPCN) to fully capture the intrinsic object details for accurate FSS. DPCN belongs to prototype-based methods yet with several elegant extensions and merits. Specifically, we first propose a dynamic convolution module (DCM) to achieve more adequate interaction between support and query features, thus leading to more accurate prediction for the query objects. As in Fig. 1(b), we leverage three dynamic kernels, i.e., a square kernel and two asymmetric kernels, generated from the support foreground features. Then three convolution operations are employed in parallel onto the query features using these dynamic kernels. This interaction strategy is simple yet important to comprehensively tackle large object variations (e.g., appearance and scales) and can capture the intrinsic object details. Intuitively, the square kernel is capable of capturing the main information of an object (e.g., main body of the plant in Fig. 1(b)); By contrast, asymmetric kernels (i.e., kernel with size d × 1 or 1 × d aim to capture the subtle object details, e.g., leaves in Fig. 1(b). As such, DPCN equipped with DCM can better handle the intrinsic object details using an extremely simple way

PROPOSED SYSTEM ADAVANTAGES:

1.HIGH ACCURACY

2.HIGH EFFICIENCY

SYSTEM REQUIREMENTS
SOFTWARE REQUIREMENTS:
• Programming Language : Python
• Font End Technologies : TKInter/Web(HTML,CSS,JS)
• IDE : Jupyter/Spyder/VS Code
• Operating System : Windows 08/10

HARDWARE REQUIREMENTS:

 Processor : Core I3
 RAM Capacity : 2 GB
 Hard Disk : 250 GB
 Monitor : 15″ Color
 Mouse : 2 or 3 Button Mouse
 Key Board : Windows 08/10