Hello everyone, this is programmer Wan Feng actively working on various AI projects.
🔗 Project Links:
- GitHub: https://github.com/openclaw/openclaw (248k+ Star) (Domestic Mirror)
- AtomGit: https://atomgit.com/openclaw/openclaw
- PyPI:
pip install openclaw(500w+ Downloads)
📚 AI Programming Learning Portal:
- "Fluent Python" 20 Lectures Advanced Programming Practice
- Python Basics Course
- Python Data Analysis Course
- CPython Design and Implementation 20 Lectures
💬 Discussion Group:
- WeChat: Scan to add friend
- Weibo: @ProgrammerWanFeng
- Zhihu: @ProgrammerWanFeng
- BiliBili: Python Automated Office Community
🦞 The Day I Received Tencent's Call, I Thought It Was a Scam
On March 10th, I received a call from Shenzhen.
"Hello, we're from Tencent Cloud Industrial Cooperation Department, we want to talk to you about free deployment of AI lobster inspection system..."
My first reaction was: Scam?
Free? A big company like Tencent would install a system for free for a small aquatic farm like ours?
But the other party was very detailed:
- Equipment they provide
- Installation they handle
- Training they conduct
- First three months cloud service fees also waived
The only requirement: Allow them to use our factory as a case study for promotion, and cooperate to collect some desensitized data.
I hesitated for two days, and finally decided to try. Anyway, it's free. The worst case is wasting a few days.
📋 Installation Process: Much Simpler Than I Thought
On March 12th, Tencent's technical team came with 3 people.
I thought there would be major construction, but:
9:00 AM - On-site survey
- Examined our sorting assembly line
- Measured installation locations
- Confirmed power and network interfaces
11:00 AM - Equipment installation
- Installed 2 industrial cameras
- Deployed light system
- Connected edge computing box (just a small host)
2:00 PM - System debugging
- Collected sample images of 500 lobsters
- Did initial model calibration
- Tested inspection process
5:00 PM - Personnel training
- Taught us how to operate the backend
- Taught us how to view data dashboard
- Taught us how to handle exceptions
Total time: 8 hours
The next morning, the system was officially operational.
💰 First Week Data: I Was Amazed
Our factory is not large, daily processing volume is about 30,000 lobsters.
Before using manual inspection:
- Needed 6 workers in shifts
- 300 yuan per person per day
- Daily labor cost: 1,800 yuan
- Monthly labor cost: 54,000 yuan
First week data after system operation:
| Metric | Expected | Actual |
|---|---|---|
| Inspection speed | 150 units/min | 187 units/min |
| Accuracy | 95% | 97.8% |
| Misjudgment rate | <3% | 1.4% |
| System stability | 99% | 99.7% |
The best part: Now only need 2 people for re-inspection and packaging, the other 4 people can do other work.
🎯 Second Week: Discovered New Problems
System operation was smooth, but also exposed some issues.
Problem 1: When small and large lobsters are inspected together, accuracy decreases
Solution: After communicating with Tencent team, they remotely updated the model, adding size pre-classification.
Problem 2: Large light changes at night, occasional misjudgments
Solution: Adjusted brightness auto-regulation parameters of light system.
Problem 3: Workers have resistance to new system
This is the funniest one. An old employee told me: "Wan Feng, will this machine take our jobs?"
I said: "No, it just helps you do the heaviest work. You do re-inspection and packaging, it's easier work."
And, because efficiency improved, we received more orders, actually needed to hire new people.
📊 One Month Later Cost Comparison
Now the system has been running for a full month, let me do a complete accounting.
Before (Pure Manual)
1 | Labor cost: 6 people × 300 yuan/day × 30 days = 54,000 yuan |
Now (AI + Manual)
1 | Labor cost: 2 people × 300 yuan/day × 30 days = 18,000 yuan |
Monthly savings: 48,000 yuan
Annual savings: 576,000 yuan
