Research: AI Based Farm Fish Disease Detection System to Help Micro and Small Fish Farmers

I. INTRODUCTION

One of the major challenges facing the World today is meeting and securing the food and nutritional needs of the teeming billions. In the backdrop of ever increasing population, this challenge turns into a complicated one. World population, which is around 7.6 billion today, is expected to increase to 9.2 billion by 2050 [17]. This situation, if not explicitly handled, is bound to give rise to huge gap between demand and supply of food. Food production needs to be increased by about 60-70 percent from the current level.[6],[7] Achieving this kind of growth, without further aggravating the load on the resources such as land, irrigation facilities and adversely impacting the environment, is a big challenge. The Food and Agricultural Organization of the United Nations (FAO), chalks out the road map for ensuring Food security, improving nutrition, to defeat hunger and making available a healthy diet for the entire population of the World. Every member Nation is encouraged to work towards this common goal. [6]

The goal committed by the World is to end hunger, food insecurity and all forms of malnutrition by 2030. This commitment calls for transformative approach encompassing efficient, resilient, inclusive and sustainable food production, storage and supply chain management. In a bid to explore all possibilities to meet the above challenge all possible avenues of food production, both conventional and non conventional needs to be considered. It is widely accepted fact that ‘Aquaculture’ can play a significant role in achieving food security.[6] Aquatic animals are highly nutritious and are the cheapest animal food sources, which serves as the valuable supplement in diet by providing essential vitamins, minerals, Omega 3 fatty acids apart from being rich source of protein. Among the animal-based food basket, farmed fish requires the lowest land use and less resource intensive, as compared to Poultry, Beef, sheep and goat meat. It is estimated that Aquaculture production needs to be more than doubled meet the projected demand in fish consumption[21]. This is envisaged to be achieved by addressing fish farms’ current environmental challenges and leveraging Technology to meet many of the challenges affecting the Aquaculture Industry.[15] Presently fish contributes to 17 percent of the animal-based protein. It is expected to rise by 58 percent. Currently there is a widespread trend in marine overfishing, which has caused marine fish depletion all around the World. The expected rise in the demand for fish and marine fish depletion, calls for giving high importance to fish farming, especially inland Aquaculture or fresh water aquaculture.[15] Aquaculture is playing a prominent role in developing countries in national economic development and global food supply. FAO has also identified the potentiality of Aquaculture sector in contributing to economic development of developing nations as well as contribution towards food basket of the world to ensure food security.[6]

Even in India this thinking has been echoed back by the policy makers and the Government. This is well reflected in the fact that in 2020 the Government has created a separate ministry of animal husbandry, Livestock and fisheries. National fisheries policy Draft 2020 has proposed a provision of Rs.20,500 crores, for the development of fisheries, which include inland fisheries as well as aquaculture.

Currently, India ranks third position in the world in fisheries and second position in aquaculture. Indian fresh water inland aquaculture Industry contributes about 55% of the total Gross value of Fish production. Fresh water aquaculture covers about 2.36 million hectares, Pond and tank aquaculture covers an area of 1.07 million hectares, Other resource such as Reservoirs (3.15 million hectares), Upland lakes (0.72 million hectares) and Canals (0.12 million KM) contributes towards the rest of the inland fisheries landscape.

II. EXISTING DATA SCIENCE APPLICATIONS IN THE INDIAN AGRICULTURE SECTOR:

Artificial Intelligence has already been playing crucial role in many areas of fish farming. Artificial Intelligence is being leveraged by the Global Aquaculture Industry to increase efficiency in production as well as moving towards sustainability. The increased importance given on reduction of load on resource as well as the environmental impact has give rise to adoption of modern improved methods of fish farming, replacing the traditional methods of fish farming. In the traditional method, which was hitherto followed in India, fish farming was done in natural habitats like ponds, lakes and canals, without creating any controlled environment. Because of this method it was difficult to implement advance technological solutions in this sector. [14] Thrust given by FAO & commitments made by the member Nations have resulting in wide scale adoption of modern controlled methods of fish farming both inland and marine, which includes Biofloc[3] system and Cage culture. This radical change in the Aquaculture Industry has resulted in huge scope for application of advanced technological solutions to monitor and control various aspects of the fish farming. Underwater camera with suitable lighting systems, IOT, communication networks and AI based applications are beginning to be accepted widely. Some of the application areas where AI is being leverage is briefly discussed below:

III. REDUCING FISH FOOD WASTAGE:

Fish food is the major component of the overall cost of fish farming. Artificial Intelligence can be leveraged, effectively to optimize fish feeding to reduce cost of feeding as well as reduce wastage being released in to open environment. In this area of fish farming[20], Artificial Intelligence applications are fast replacing traditional methods which solely relied on people. Efficient food pellet dispersion, using Artificial Intelligence systems economizes on the consumption of food pellets, while at the same time reducing the impact on the environment, by eliminating wastage to a large extent. This leads to substantial cost saving to the farmers. [4]

Goals adopted by such Artificial Intelligence systems include:[5]

1. Systems that provide farmers empirical and objective guidance on how much and when to feed.
2. Systems that use sensors to detect hunger level in fish for controlling the food dispensers, which release just adequate amount of food pellets to prevent both under feeding as well as wastage.
3. Data driven food dispensers, with food feeding schedules. These systems include learning systems which improve as more and more data is gathered and underlying model is optimized.

IV. DISEASE DETECTION:

Traditional systems of disease detection call for continuous attention of the farmers to keep a frequent watch on the fishes to detect the onset of diseases in the fish and then to take proper remedial action to treat the disease appropriately, so that the disease does not spread to the whole stock, which may result in huge loss. [9] [10] Artificial Intelligence based disease detection systems [10] enable timely intimation to the concerned farmer, without any person being in constant watch over the fish farm.[4] This hugely impacts the work-life balance of the farmer in a positive manner. Apart from this a well trained system would be more accurate than an average farmer, who may fail to detect[19] in time, the onset of any disease because of either ignorance or fatigue. Such systems can also be integrated with treatment triggering systems, which will automate disease management of the fish farm holistically. Methods adopted for disease detection can include the following approaches:

1. Image based detection[8] system which relies on the images gathered from a given fish pen to collect images of the fishes from time to time and then score to classify, based on the pre-trained Machine learning models to detect diseases. Such information can automatically be passed on to the stake holders for taking remedial action.[4]
2. In the advanced systems video based detection can be adopted to detect the diseases and then to automatically trigger dispensation of required chemicals to regulate the spread of disease in the pen, as well as the surrounding pens situated in the same habitat.
3. Integrating motion detection along with image processing may constitute more advanced system. Sensors may also be utilized to gather information related to onset of disease. These kinds of systems also provide scope for adopting treatment triggering approaches as soon as any disease is detected, while at the same time intimating all the concerned stake holders.

V. IN SITU PARAMETER MONITORING IN THE AQUATIC ECOSYSTEM:

For a healthy and optimal yield of fishes it is very essential to monitor various aspects of the given fish farm ecosystem, like water quality, oxygen, Nitrogen and phosphorous levels in the water, presence of other types pollutants, presence of pro-biotic and disease causing bacteria etc. While optimal oxygen level is beneficial, high levels of Nitrogen and phosphorous is known to trigger algal and cyanobacterial bloom leading to reduced oxygen content in water, which is not conducive to health of the farm fishes.

Artificial intelligence along with sensor based devices can go a long way in fulfilling these functions without elaborate intervention of trained professionals.[4] Such approaches are being adopted to improve the quality of environment in which fishes are bred and farmed. Even in India some companies are doing some good work in this area.

A. Yield Management:

Presently, fish farmers assess the growth of their fishes through personalized intuitive methods of experienced farmers. This may not be the best scientific method. Adopting methods based on Artificial Intelligence, farmers can easily control the feeding cycles and schedules and quantum so as to be able to predict exactly when their fish will be ready for harvesting so as gain optimum yield. Such systems are being adopted both in India as well as global level. Non intrusive methods of assessing the weight of each fish are available, which is very helpful to the fish farmers.

B. Price movement data:

Even after putting in all the efforts fish farmers are not assured of remunerative price for their product in the market. Fish being a perishable item it has short shelf life after harvesting[17]. There are many risk factors beyond the control of the fish farmers that may affect the realization of good price for their harvest. To mitigate some of the risks involved in fish farming Artificial Intelligence can play a crucial role.[4] Practically Artificial Intelligence is being leveraged to collect market data and predict in advance the possible gate price of fishes, so that farmers can decide on the schedule and quantum of fishes to be harvested to fetch the maximum returns on their investment. Based on the historical data the fish farmers can be in a situation to make important decision as to when to start the breeding process and how much quantity to plan for periodical yield, based on the information provided by the Artificial Intelligence systems[17].

C. Application proposed to be developed by us:

Though there are some players providing all of the services mentioned above it is noticed that Micro and small Farmers are a marginalized segment with insufficient financial capacity to subscribe to the services presently available. So we are intending to provide Niche service specifically targeted towards these segments. Our goal is to develop a system based on Artificial Intelligence to detect Fish disease. The system by leveraging image processing technology [11] will be able to detect disease as and when afflicted within a reasonable time and intimate all stake holders including the concerned Fish Farmer. In turn the Farmer can take appropriate step to contain, the spread of disease. This will help the farmer from incurring heavy losses in case of affliction of any of the Fish disease. [4] The salient feature[18] of our solution is that the service is provided at a very low cost, keeping in mind the financial status of Micro and Small Fish Farmers, prevailing in India. Else, the Fish Farmers in this segment will not be able to make use of such Technology. Keeping the cost low is intended to be achieved even without expecting incentives from Government agencies. In case of receipt of incentives the same can be utilized to provide additional extended facilities for efficient and economical management of the Fish farms.

Our proposed system will consist of Underwater camera with suitable system of Lighting to obtain quality images from time to time, which will be passed on to Azure Cognitive Services on cloud, for processing, scoring and classifying based on the Trained model that has been put into place by us initially by training with data obtained from Government and other agencies dealing in the field. Once classified and if fishes are found to be afflicted by disease, a suitable, predetermined message will be passed on to the stake holders, including the concerned Fish Farmer. This will enable the concerned Fish Farmer to take immediate steps to contain the further spread of diseases both by the Fish Farmers and all the other concerned stake holders. By adopting this system Fish Farmers are expected to be protected from incurring of losses due to the onset of any of the Fish diseases.

Our proposed Business model will work on Multi Tenancy Subscription Model. The Fish Farmers will not be required make any kind of initial investment to benefit from the system. All the benefits will be available through a low monthly subscription.

The financial details and Logical set up, Process flow and Business model concepts are depicted in the table and diagrams appended at Appendix 1 to 3 below.

VI. APPENDIX

Appendix – 1: Business Plan (Financials): Microsoft Azure Services - Estimate (Region: Central India)

Service type	Description	Estimated monthly cost	Estimated upfront cost
Azure Cognitive Services	Computer Vision, Pay as you go, S1: 1 thousand Tag, Face, Get Thumbnail Colour, Image Type transactions; 0 thousand OCR (printed), Adult, Celebrity and Landmark transactions; 0 thousand Describe and OCR (handwriting) transactions.	₹ 74.40	₹ 0.00
Storage Accounts	Block Blob Storage, General Purpose V2, LRS Redundancy, Hot Access Tier, 100 GB Capacity - Pay as you go, 10 x 10,000 Write operations, 10 x 10,000 List and Create Container Operations, 10 x 10,000 Read operations, 1,00,000 Archive High Priority Read, 1 x 10,000 Other operations. 1,000 GB Data Retrieval, 1,000 GB Archive High Priority Retrieval, 1,000 GB Data Write	₹ 234.23	₹ 0.00
Azure Functions	Consumption tier, 128 MB memory, 100 milliseconds execution time, 10,000 executions/month	₹ 0.00	₹ 0.00
Azure Cosmos DB	Standard provisioned throughput (manual), Single Region Write (Single-Master) - Central India (Write Region); 400 RU/s x 729 Hours; 0 GB transactional storage, 2 copies of periodic backup storage; Dedicated Gateway not enabled	₹ 1,974.18	₹ 0.00
Support	Support	₹ 0.00	₹ 0.00
Total		₹ 2,282.81	₹ 0.00

The above cost is factored for up to 100 customers (or 500 pens/cages). Operations above 100 customers (500 pens) would proportionately increase some costs. Overall would work out to be cheaper than indicated above. Cost per pen works out to Rs.22.81 excluding manpower costs. This will enable the company to keep the subscription rates to the most affordable level.

Appendix – 2: Solution At A Glance – Technical architecture