\u2728Common Object Detector Architecture you should be familiar with:
— AI Fast Track (37/45) (@ai_fast_track) October 27, 2021
\U0001f4cc Common object detectors are divided into One-Stage Detectors (OSD), and Two-Stage Detectors (TSD)
\U0001f4cc Both OSD and TSD can be either anchor-based (relying on anchor boxes) or anchor-free pic.twitter.com/TGxS0IbE8g
4 Feature Pyramid Network (FPN) Design you should know:
— AI Fast Track (37/45) (@ai_fast_track) November 15, 2021
FPN, PANet, NAS-FPN, and BiFPN
\U0001f4cc (a) FPN uses a top-down pathway to fuse multi-scale features from level 3 to 7 (P3 - P7);
\U0001f4cc (b) PANet adds an additional bottom-up pathway on top of FPN; pic.twitter.com/k22vo6Df4L
\U0001f9d07 things you should know about the Focal Loss:
— AI Fast Track (37/45) (@ai_fast_track) October 25, 2021
\U0001f4cc It was introduced in the RetinaNet paper to address the foreground-background class imbalance encountered during training of dense detectors (one-stage detectors)
... pic.twitter.com/NaXzh80Etd
FCOS is an an anchor-free object detector.
— AI Fast Track (37/45) (@ai_fast_track) November 20, 2021
It was one of first competitors of anchor-based single/two stage object detectors.
Understanding FCOS will help understanding other model inspired by FCOS.
Summary ...\U0001f447 pic.twitter.com/aPFJ0h1olz
\U0001f947YOLOX beat YOLOv5. They won the 1st Place on Streaming Perception Challenge (Workshop on Autonomous Driving at CVPR 2021).
— AI Fast Track (37/45) (@ai_fast_track) October 21, 2021
I briefly explain how they created it, here below\U0001f447
YOLOX will be soon supported in our IceVision Framework (Repo: https://t.co/wNQen4X65M) pic.twitter.com/CBGE1QbCNT
Let's see how many lines of code you need to create a YOLOV5 or YOLOX model in IceVision?
— AI Fast Track (37/45) (@ai_fast_track) November 3, 2021
Dozens? Hundreds? Your Guess
2 lines!\U0001f92f
and swapping models is as easy as changing a few words pic.twitter.com/4YZFc210X3
\u2747VFNet: A very interesting model that isn\u2019t under the radar. You should give it a try :)
— AI Fast Track (37/45) (@ai_fast_track) November 17, 2021
VariFocalNet: An IoU-aware Dense Object Detector
\U0001f9ca Background:
\U0001f4cc Accurately ranking candidate detections is crucial for dense object detectors to achieve high performance
... pic.twitter.com/MugxaA9Cxf
YOLO Real-Time (YOLO-ReT) architecture targets edge devices.
— AI Fast Track (37/45) (@ai_fast_track) November 2, 2021
It achieves 68.75 mAP on Pascal VOC and 34.91 mAP on COCO using MobileNetV2\xd70.75 backbone.
Here is a brief description of the YOLO-ReT \U0001f447 pic.twitter.com/md8DjWinjO
This illustration shows you both the similarities and the differences between some popular Object Detection models.
— AI Fast Track (37/45) (@ai_fast_track) November 12, 2021
I shared more detailed information about some of those models here. Checkout out my threads.
All those models are supported in IceVision. pic.twitter.com/Yhp00v3upJ
\U0001f947 FCOS3D won the 1st place out of all the vision-only methods in the nuScenes 3D Detection Challenge of NeurIPS 2020.
— AI Fast Track (37/45) (@ai_fast_track) November 4, 2021
Here is a brief description:
\U0001f4cc FCOS3D is a monocular 3D object detector
\U0001f4cc It\u2019s an anchor-free model based on FCOS (2D) counterpart pic.twitter.com/Muzah3YIXw
The Generalized Intersection over Union (GIoU) can be used both as a metric and a loss function
— AI Fast Track (37/45) (@ai_fast_track) November 11, 2021
Try it out and see how your model performance will improve
... \U0001f447 pic.twitter.com/Ee982mvkbB
What is the Average Precision (AP), mean AP (mAP), and COCO Metric?
— AI Fast Track (37/45) (@ai_fast_track) November 9, 2021
The COCO metric is widely used in evaluating and comparing model performances. pic.twitter.com/QiVUjIh2uJ
I\u2019m at Day 23 of my 30 posts (on Object Detection) in 30 days challenge
— AI Fast Track (37/45) (@ai_fast_track) November 23, 2021
I gathered 12 visual summaries on OD Modeling \U0001f381
A lot of people find those posts helpful, follow @ai_fast_track to catch the upcoming posts, and give this tweet a quick retweet \U0001f64f
Summary of summaries\U0001f447 pic.twitter.com/DMD0owByhS
1. Let's start option selling learning.
— Mitesh Patel (@Mitesh_Engr) February 10, 2019
Strangle selling. ( I am doing mostly in weekly Bank Nifty)
When to sell? When VIX is below 15
Assume spot is at 27500
Sell 27100 PE & 27900 CE
say premium for both 50-50
If bank nifty will move in narrow range u will get profit from both.
Few are selling 20-25 Rs positional option selling course.
— Mitesh Patel (@Mitesh_Engr) November 3, 2019
Nothing big deal in that.
For selling weekly option just identify last week low and high.
Now from that low and high keep 1-1.5% distance from strike.
And sell option on both side.
1/n
Sold 29200 put and 30500 call
— Mitesh Patel (@Mitesh_Engr) April 12, 2019
Used 20% capital@44 each
Already giving more than 2% return in a week. Now I will prefer to sell 32500 call at 74 to make it strangle in equal ratio.
— Mitesh Patel (@Mitesh_Engr) February 7, 2020
To all. This is free learning for you. How to play option to make consistent return.
Stay tuned and learn it here free of cost. https://t.co/7J7LC86oW0
