Object Detection — Most Important “B” Detectors and Use-cases for Smart Cities

Author(s): Mihir Rajput

Originally published on Towards AI.

Computer Vision

Let’s dive in. In this article will cover the following objects detection.

1. Bus/Heavy Vehicle Detection
2. Birds Detection
3. Baseball bat Detection
4. Bicycle Detection
5. Bottle Detection (The world environment day special!)

These all the objects are trained individually to be detected. The trained models/weights are open-sourced for non-commercial use. You can contact me via email, through GitHub, or LinkedIn for the same. Also, I have created processed videos and uploaded them on YouTube that represent the models’ performance on unseen data.

Bus/Heavy Vehicle Detection

Importance/Facts

Buses and Heavy vehicles are common in any country, city, or area. Buses are used by so many people to travel within a city or state. Heavy vehicles are used by many companies to transfer goods and stuff. Hence this enables the opportunities for AI applications related to monitoring, analysis, and automation.

Usecases

1) Bus monitoring/management system for the government, where the government can keep track of using CCTV cameras.

They can,

– Count the ingoing/outgoing buses on the bus depot.

– Verify the bus’s arrival timings.

– Track the bus arrival frequency for the particular bus station.

This kind of generated data can help the government to solve many problems.

2) Vehicle type(heavy or small) is a major aspect of the toll-tax and highway-related services, so maybe they can use this somehow by combining CCTV footage and computer vision to solve their problems.

3) There are some areas or bridges where heavy vehicles are prohibited. We can use heavy vehicle detection to monitor those areas.

Output

bus/heavy vehicle detection

Birds Detection

Importance/Facts

Fact-1,

A bird strike — sometimes called birdstrike, bird ingestion (for an engine), bird hit, or bird aircraft strike hazard (BASH) — is a collision between an airborne animal (usually a bird or bat) and a manmade vehicle, usually an aircraft. The term is also used for bird deaths resulting from collisions with structures such as power lines, towers, and wind turbines (see Bird–skyscraper collisions and Towerkill).

These cause annual damages that have been estimated at $400 million[3] within the United States alone and up to $1.2 billion to commercial aircraft worldwide.

Fact-2,

Tens of millions of flamingos, storks, pelicans, and other migratory birds are being killed across the world when they fly into power lines, according to a new study.

Now that we know this use case is important. Let’s explore some of its applications and outputs.

Fact-3,

Crop loss in horticulture due to birds is an on-going and increasing cost to growers. Estimates of damage vary but are generally reported in the research literature as 30% to 35% of small berry production, 7% for wine and table grapes, 13% for apples and pears, 16% for stone fruits, and 22% in the nut crops.1 This includes whole fruit being consumed. Fruit knocked off bushes or canes, and unsalable fruit (pecks, holes, slashes). These estimates are consistent with the OFVGA grower survey conducted in 2012, which reported 35% losses in berry crops, 10% loss for apples, 15% for tender fruit, and 7% for grapes.

Usecases

1) Bird caretakers, people who pet or sell birds(especially white pigeon caretakers because in India they allow pigeons to fly for some distance), can keep track of birds, count them and monitor them easily.

2) At the airport runway and other airplane landing and take-off areas, we can place a bird monitoring system to prevent major accidents caused by bird strikes.

3) We can place a monitoring system near electrical power plants, huge mobile towers, or any other areas which cause huge bird deaths.

4) Monitoring systems in crop yards and farms could help farmers to avoid crop and vegetable damage.

Well, you might be wondering how to prevent these accidents and how to keep birds away from there areas?

Answer to that is by combining detection +alarm triggering. When there is detection, you can trigger the alarms which make certain noises to make birds fly away!

Output

Baseball bat Detection

Importance/Facts

“Baseball bats, although meant for recreational use, are commonly used as assault weapons. Here in the UK, assault is more likely to occur with body parts only: however, a trend for using baseball bats has been observed both by emergency departments and the police. The bat is an easily acquired weapon, a simple wooden one being available for £15 at a sports store. At present, there are no restrictions on the purchase of these bats that represent a major cause of morbidity and occasionally mortality when used in an assault.”

The injuries sustained fell into three broad categories:

• Facial trauma — 6 (30%) cases: 3 nasal fractures, 2 malar fractures, 1 zygomatic arch fracture.
• Head injury — 8 (40%) cases: mostly minor injuries and scalp lacerations, no skull fractures or intracranial injury in this study.
• Extremity trauma — 6 (30%) cases: 5 soft‐tissue injuries to the upper limb, 1 tibia fracture.

The patient outcomes were recorded as follows:

• Discharged no review: 4 (20%) cases.
• Discharged with clinic (emergency department of maxillofacial) review: 8 (40%) cases.
• Admitted to hospital: 8 (40%) cases.

Sources:

1. Groleau G A, Tso K L, Olshaker J S.et al Baseball bat assault injuries. Trauma J 199334366–372. [PubMed] [Google Scholar]

2. Ord R A, Benian R M. Baseball bat injuries to the maxillofacial region caused by an assault. Oral Maxillofac Surg J 199553514–517. [PubMed] [Google Scholar]

3. Berlet A-C, Talenti D P, Carroll S F.et al The baseball bat: a popular mechanism of urban injury. Trauma J 199233167–170. [PubMed] [Google Scholar]

Usecases

1. This object detection can be used in public areas, events, or parties where the organizers or government wants to prevent these kinds of objects from entering.
2. Baseball detection is also useful to detect violence from security cameras. Of course, this will require some extra efforts, we may need to embed this with any activity recognition algorithm.

Output

Bicycle Detection

Importance/Facts

Well recently, bicycle riding is a new trend, and many people are adapting it. Some CEOs, actors, businessmen, office workers prefer bicycles over other vehicles. This preference is mainly for health and environmental contribution.

Where I live, in Ahmedabad, there is a project for bicycle sharing and renting. Take a look here https://mybyk.in/.

As these types of “Bicycle” related projects are taking place, it’s getting trending again to ride a bicycle instead of other vehicles.

Well, as things get trendy, always take benefits out of it. As an AI engineer, it’s easy for me to pick any trending opportunities and co-relate/combine it with AI and come up with some use-cases and solutions.

Usecases

1. Projects like bicycle sharing can use bicycle detection to manage, count, and secure their bicycles automatically using CCTV cameras.
2. Bicycle racing or any other related events can use bicycle detection using cameras to get notifications(To know where the participants have reached because the track sometimes is very long), interesting automated insights like how many participants have reached/covered certain milestones.
3. Bicycle detection is also useful to automate the monitoring for some places where bicycles are prohibited(like National Highways) or only the bicycles are allowed (Like Riverfronts or gardens).
4. Also, this could be useful in smart cities to get interesting insights and information about how many people use bicycles or what’s the frequency of riding bicycles. Which area has the most number of bicycles and which has the least number of bicycles?

Output

Bottle Detection

Importance/Facts

Bottles are one of the most essential parts of our day to day life. Everyone uses them. Plastic bottles, steel or aluminum bottles, water bottles, beverages bottles, in short, the bottles are everywhere!

On the other hand, plastic bottles are affecting the environment rapidly. Its causing damage to lands and oceans. Here are some examples/facts,

• Plastic bottles require up to 700 years to dissolve.
• 90% of the cost of bottled water is the bottle itself.
• 80% of plastic bottles never get recycled.
• 38 million plastic bottles go to landfill each year in America alone.
• 24,000,000 liters of oil are needed to produce these billions of plastic bottles.
• The average American consumes 167 bottles of water per year.
• Bottling water and shipping transport is the least energy-efficient method of water supply in the history of mankind.
• Bottled water is the second most popular beverage in the United States.

source: https://myequa.com/blogs/blog/89626311-facts-about-plastic-bottles

As an environment-friendly person, I would suggest that plastic bottles should not be used or used with limitations.

To do so, automated monitoring would be required. That’s the important reason I chose to train the bottle detection algorithm on this world environment day!

Usecases

• Beaches, lakes, gardens, cinema halls, some streets where bottles are not allowed, bottle detection, can be used to monitor these areas automatically using CCTV cameras.
• Bottle manufacturers can use this for counting and monitoring purposes.
• Automated or ROBO hotels(which uses automated serving using robots) can use bottle detection to verify if all the essentials(water bottles primarily) are available on the table or not.

Output

In the ending notes, I encourage you to use these use-cases and outputs, build the awesome applications, integrate them in smart city projects, enhance your knowledge and skills.

Any feedback, suggestions would be really appreciated.

You can connect with me via LinkedIn or GitHub.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI