japan ai ficha.jp company all GitHub code

#1
The corporate code platform GitHub of the well‑known Japanese AI development company ficha.jp contains all of the company’s development code as well as customer‑specific project code. 
Hidden Content
You must register or login to view this content.


# **English Translation**
## **1. SDK Core Platforms (Most Critical)**
### **sdk2 ⭐⭐⭐⭐⭐**  
**Role:** Second‑generation SDK platform; the company’s core technical framework  
**Derived projects:**  
- sdk2-desayface — Desay face-recognition version  
- sdk2-for-desay — Desay customized version  
- sdk2-for-koito — Koito customized version  
- sdk2-JVC-adas — JVC ADAS version  
- sdk2-JVC-ADAS-Android — JVC Android version  
- sdk2-translog-B2C — B2C transaction‑log version  
**Features:**  
- Multiple customer‑specific branches  
- Cross‑platform support (Android, embedded systems)  
- Represents the company’s “productized” core
### **SDK4 ⭐⭐⭐⭐⭐**  
**Role:** Fourth‑generation SDK platform (latest version)  
**Features:** Iterative technical upgrade, likely integrating experience from sdk2
### **AIFramework ⭐⭐⭐⭐⭐**  
**Role:** Underlying AI framework; foundation of all AI functionalities  
**Status:** Comparable to an “operating‑system‑level” component
---
## **2. ADAS Core Algorithms (Highly Productized)**
### **FichaDet ⭐⭐⭐⭐⭐**  
**Role:** Ficha’s proprietary object detector  
**Features:**  
- Named after the company (Ficha = brand)  
- Core intellectual property  
- Likely the foundation of many customer projects
### **LaneDet / LaneDetection ⭐⭐⭐⭐⭐**  
**Role:** Core lane‑detection algorithm  
**Importance:** One of the fundamental ADAS functions  
**Usage:** Used in nearly all ADAS customer projects
### **VP_Detection / CalcVp ⭐⭐⭐⭐**  
**Role:** Vanishing‑point detection and computation  
**Importance:** Key technology for lane and road‑boundary detection
### **TrafficSignLightClassification ⭐⭐⭐⭐**  
**Role:** Traffic sign & traffic light classification  
**Importance:** Standard ADAS functionality
### **FC_tracker ⭐⭐⭐⭐**  
**Role:** Full‑circle tracker  
**Features:** Likely a core multi‑object tracking technology
---
## **3. DMS Core Algorithms (Independent Product Line)**
### **DMS Neural Network Series ⭐⭐⭐⭐⭐**  
- DMS_InsightFace — Face recognition (based on open‑source InsightFace)  
- DMS_BHNet — Behavior detection network  
- DMS_BNet — Base network  
- DMS_ETNet — Eye‑tracking network  
- DMS_EyeNet — Eye‑detection network  
- DMS_LMNet — Landmark detection network  
**Features:**  
- Complete DMS technology stack  
- Modular design (face, eyes, behavior separated)  
- Indicates DMS is an independent business line
### **FACE-EYE / FACE-EYE2 ⭐⭐⭐⭐⭐**  
**Role:** Joint face + eye detection  
**Versioning:** FACE‑EYE2 is the improved version
### **GazeEstimation / HeadPoseEstimation ⭐⭐⭐⭐**  
**Role:** Gaze estimation + head‑pose estimation  
**Importance:** Core functions for driver attention monitoring
### **drowsiness ⭐⭐⭐⭐**  
**Role:** Fatigue detection  
**Importance:** Critical safety function in DMS
### **dms-demo-library ⭐⭐⭐⭐**  
**Role:** DMS demo library (for external demos / customer presentations)
---
## **4. OCR Core Projects (New Business Direction)**
### **pure_ocr_sdk_libtorch ⭐⭐⭐⭐⭐**  
**Role:** Pure OCR SDK based on LibTorch  
**Features:**  
- “Pure” indicates an independent SDK  
- LibTorch enables production‑grade deployment  
- Represents the latest OCR core
### **table_sdk ⭐⭐⭐⭐**  
**Role:** Table‑recognition SDK  
**Importance:** Advanced OCR capability (invoices, forms)
### **OCR_API / OCR_SERVER / OCR_WEB ⭐⭐⭐⭐**  
**Role:** Three deployment forms of OCR  
**Indicates:** Productized with multiple access methods
### **drawing-llm / insurance-llm ⭐⭐⭐⭐**  
**Role:**  
- drawing-llm: Blueprint / engineering drawing analysis  
- insurance-llm: Insurance‑related document analysis  
**Features:**  
- Introduces LLM technology  
- Represents new technical directions
### **mold_design_drawing_analysis ⭐⭐⭐⭐**  
**Role:** Mold design drawing analysis  
**Indicates:** OCR expansion into industrial applications
---
## **5. Inference Engines (Deployment Core)**
### **ncnn Series ⭐⭐⭐⭐⭐**  
- ncnn-classifier — NCNN classifier  
- ncnn_int1178 — NCNN integer quantization (likely mixed int8/int11)  
- ncnn_tester — NCNN testing tools  
**Importance:**  
- NCNN is a mainstream inference engine for mobile/embedded  
- Quantization (int1178) indicates deep optimization
### **MNN_int1178 ⭐⭐⭐⭐**  
**Role:** Alibaba MNN inference engine  
**Indicates:** Multi‑engine deployment strategy
### **pytorch2ncnn ⭐⭐⭐⭐**  
**Role:** Model conversion tool  
**Importance:** Bridge from training to deployment
---
## **6. Major Customer‑Specific Projects**
### **Bosch Series ⭐⭐⭐⭐⭐**  
- Bosch-SDK — Bosch SDK (global Tier‑1 giant)  
- Bosch-ROS — ROS integration (Bosch requires ROS architecture)  
- Bosch-tcpip — TCP/IP communication protocol  
- Adrien_Bosch_V4H — V4H platform (Renesas high‑end chip)  
**Importance:**  
- Bosch = world’s largest Tier‑1 supplier  
- Independent SDK indicates large project scale  
- V4H = high‑end ADAS solution
### **Clarion Series ⭐⭐⭐⭐**  
- Clarion — Main project (Hitachi Group)  
- Clarion_RearAEB_H3 — Rear AEB (Automatic Emergency Braking)  
**Importance:**  
- Hitachi subsidiary; major Japanese automotive supplier  
- AEB = safety‑critical function
### **JVC Series ⭐⭐⭐⭐**  
- HybridDL_JVC_ADAS_2020 — Hybrid deep‑learning ADAS  
- sdk2-JVC-adas — JVC ADAS SDK  
**Features:**  
- HybridDL likely combines traditional + deep learning  
- “2020” indicates long‑term cooperation
### **Toyota Series ⭐⭐⭐⭐**  
- ToyotaImageRecog — Toyota image recognition  
- ToyotaRoadCrackDetection — Road‑crack detection  
**Importance:**  
- Toyota = world’s largest automaker  
- Road‑crack detection = specialized application
---
## **7. Hardware Platform Adaptation (Technical Barrier)**
### **Jetson Series ⭐⭐⭐⭐⭐**  
- jetson-demo — Jetson demo  
- adas-dms-jetson-demo-app — Full ADAS + DMS demo  
**Importance:**  
- NVIDIA Jetson = mainstream automotive AI platform  
- Full ADAS+DMS indicates technical maturity
### **Renesas V4H ⭐⭐⭐⭐⭐**  
- Adrien_Bosch_V4H — V4H adaptation (Bosch project)  
- bosch_communication_sample_v4h  
**Importance:**  
- V4H = Renesas high‑end ADAS chip  
- Widely used by Japanese automakers  
- High technical barrier
### **Zynq Platform ⭐⭐⭐**  
- koito-fir_in_light-zynq — Koito FIR‑in‑headlamp project  
**Specialty:**  
- FPGA platform → hardware acceleration capability  
- Infrared detection inside headlamp = innovative application
---
## **8. Tools & Frameworks (R&D Efficiency)**
### **AutoBuild ⭐⭐⭐⭐**  
**Role:** Automated build system  
**Importance:** Infrastructure for multi‑project management
### **FichaCSVtoMSCNN ⭐⭐⭐⭐**  
**Role:** Data format conversion (CSV → MSCNN training format)  
**Indicates:** Custom data‑annotation workflow
### **data-comparator ⭐⭐⭐**  
**Role:** Data comparison tool  
**Usage:** Likely for regression testing / accuracy comparison
### **model_repository ⭐⭐⭐⭐**  
**Role:** Model repository (centralized management of trained models)  
**Importance:** Model version control
---
## **Summary of Core Projects (Top 20)**
### **Platform Level (3)**  
1. sdk2 ⭐⭐⭐⭐⭐  
2. SDK4 ⭐⭐⭐⭐⭐  
3. AIFramework ⭐⭐⭐⭐⭐  
### **ADAS Core (5)**  
4. FichaDet ⭐⭐⭐⭐⭐  
5. LaneDet / LaneDetection ⭐⭐⭐⭐⭐  
6. VP_Detection ⭐⭐⭐⭐  
7. TrafficSignLightClassification ⭐⭐⭐⭐  
8. FC_tracker ⭐⭐⭐⭐  
### **DMS Core (4)**  
9. DMS_InsightFace ⭐⭐⭐⭐⭐  
10. FACE-EYE2 ⭐⭐⭐⭐⭐  
11. GazeEstimation ⭐⭐⭐⭐  
12. drowsiness ⭐⭐⭐⭐  
### **OCR Core (3)**  
13. pure_ocr_sdk_libtorch ⭐⭐⭐⭐⭐  
14. table_sdk ⭐⭐⭐⭐  
15. drawing-llm ⭐⭐⭐⭐  
### **Inference Engines (2)**  
16. ncnn_int1178 ⭐⭐⭐⭐⭐  
17. pytorch2ncnn ⭐⭐⭐⭐  
### **Customer Projects (3)**  
18. Bosch-SDK ⭐⭐⭐⭐⭐  
19. HybridDL_JVC_ADAS_2020 ⭐⭐⭐⭐  
20. Clarion_RearAEB_H3 ⭐⭐⭐⭐  
---
## **Criteria for Identifying Core Projects**
1. **Naming patterns:**  
   - Company‑named (FichaDet, FichaCSVtoMSCNN)  
   - SDK‑based names (sdk2, SDK4, Bosch-SDK)  
   - Version iterations (FACE-EYE2, SDK4)
2. **Derived branches:**  
   - sdk2 → 6 customer versions  
   - DMS → 7 specialized networks  
3. **Technical depth:**  
   - Proprietary algorithms (FichaDet, FC_tracker)  
   - Quantization optimization (int1178)  
   - Multi‑platform adaptation (Jetson, V4H, Zynq)
4. **Customer level:**  
   - Bosch (global Tier‑1 giant)  
   - Toyota (world’s largest automaker)  
   - Clarion/Hitachi (major Japanese supplier)
5. **Business completeness:**  
   - Full ADAS pipeline (detection → tracking → classification)  
   - Full DMS stack (face → eyes → behavior → fatigue)  
   - Full OCR workflow (recognition → tables → LLM)
6. **Technical forward‑looking aspects:**  
   - LLM integration (drawing‑llm, insurance‑llm)  
   - Hybrid deep learning (HybridDL)  
   - Multi‑sensor fusion (FIR infrared)
---
## **If Only 10 Projects Could Be Kept**
1. sdk2 / SDK4 — Core platforms  
2. AIFramework — AI foundation  
3. FichaDet — Proprietary detector  
4. DMS_InsightFace — DMS core  
5. pure_ocr_sdk_libtorch — OCR core  
6. ncnn_int1178 — Inference engine  
7. Bosch-SDK — Most important customer project  
8. LaneDet — ADAS fundamental  
9. GazeEstimation — Key DMS function  
10. drawing-llm — Future direction  
These represent the company’s technological moat and commercial value.
Reply
#2
what is unzip password?
Reply



Recently Browsing 2 Guest(s)