Skip to content

Instantly share code, notes, and snippets.

{
"/root/tv-process/frames/basketball/output_frame_0019.png": {
"rating": "blurry",
"qalign_aesthetic": 2.91796875,
"qalign_quality": 4.109375,
"centroid": 83,
"gpt4_caption": "In evaluating the provided image based on the focused criteria, here are my observations:\n\n1. **Subject Focus and Detail**: The players are meant to be the main subjects in this sports scene. Upon close inspection, individual players do not exhibit high clarity or crisp textures. Apparel and facial expressions are somewhat discernible, but the details are not sharp.\n\n2. **Sharpness and Definition**: The edges of the subjects (players) are not very crisp, showing a mild softness that impacts the overall sharpness. The floor markings and surrounding audience are clearer but they are not the main subject.\n\n3. **Motion Blur**: There is a slight motion blur noticeable around some players, likely due to their fast movement during the game. This contributes to a decrease in the image clarity where it is most critic
{
"/root/tv-process/frames/Friends-Funniest-Moments!-_Friends-RjpvuPAzJUw/output_frame_0063.png": {
"rating": "good",
"qalign_aesthetic": 2.611328125,
"qalign_quality": 4.4375,
"centroid": 53,
"gpt4_caption": "1. High-resolution evaluation:\n - The subjects (two women) are the primary focus.\n - Facial features, clothing textures, and hair details are distinguishable, but contours and finer details are slightly soft, particularly around the edges and hair.\n \n2. Clarity and sharpness:\n - The image shows some softness in detail and does not achieve full sharpness, particularly noticeable in the finer hair strands and elements of the clothing. \n - Although faces are recognizable, the image lacks the desired high-definition sharpness for training data, as facial features should display more distinct contrast and crispness.\n\n3. Evaluation for motion blur, lens blur:\n - There is no significant motion blur.\n - Some potential lens blur or compression artifact reduces the fi
You are an advanced visual analysis AI model. Your task is to carefully examine the provided images and determine if they meet the required quality standards for machine learning model training.
For each image, follow these steps:
1. Focus on the subject of the image at high resolution, paying close attention to small details, clarity, and blurriness
2. Ignore the parts of the image that are not in focus (i.e. the Bokeh effect)
3. Evaluate the overall clarity, sharpness, and definition of the subject. Look for crisp edges, distinct textures, and well-defined features. Check for motion blur, lens blur, or hazy/smeared appearance of the subject
The primary subject of the image must be clear, sharp, and free of any noticeable blur or loss of detail to be considered high-quality training data. Very slight blurriness is acceptable.
For each image, output one of the following quality ratings based on the above instructions:
{
"/root/tv-process/frames/Friends-Funniest-Moments!-_Friends-RjpvuPAzJUw/output_frame_0063.png": {
"rating": "good",
"qalign_aesthetic": 2.611328125,
"qalign_quality": 4.4375,
"centroid": 53,
"gpt4_caption": "In analyzing the provided image against the quality standards set for training machine learning models, the following observations have been made:\n\n1. **Clarity and Detail**: The image shows medium resolution but the details on the primary subjects, two women, aren't extremely sharp. There's a noticeable softness particularly around facial features which lack crispness. This might impact the effectiveness of training data that requires facial recognition or detailed expression analysis.\n\n2. **Motion and Lens Blur**: There's no obvious motion blur; the subjects are static. However, there is a slight blurriness, likely resulting from lens issues or post-processing that smudges finer details.\n\n3. **Overall Sharpness and Definition**: While the broader features such as clothing a
# Hello
# Yay!
# TODO maybe put units in variables
from collections import namedtuple
from collections import defaultdict
BatteryModule = namedtuple('BatteryModule', ['bank_id', 'string_id', 'module_id', 'voltage', 'temp'])
ModuleId = namedtuple('BatteryModule', ['bank_id', 'string_id', 'module_id'])
const { Socket } = require('net');
const { Duplex } = require('stream');
class JsonSocket extends Duplex {
/**
JsonSocket implements a basic wire-protocol that encodes/decodes
JavaScripts objects as JSON strings over the wire. The wire protocol
is defined as:
4 len - length of JSON body
/**
* Usage: Call BigAutorelease* a = create_autorelease_object();. If the dealloc message is not printed, this code is not running in an autorelease pool block.
*/
@interface BigAutorelease: NSObject
@property(strong, nonatomic) NSMutableArray *arr;
@end
@implementation BigAutorelease
- (instancetype)init {
NSLog(@"Make BigAutorelease");
@theicfire
theicfire / -
Created January 10, 2015 20:01
hi there
# Find sqrt(2)
# Get a graph of y^2 - 2, find the 0. Use newtons's method.
import math
def iterate_root(start, n, p):
dx = .01
dy = ((start + dx) ** p) - (start ** p)
return start - ((start**p) - n)/(dy/dx)
def get_sqrt(n):
# Q: Given a list of length n+1 with numbers from 1 to n, there will be at least one duplicate. Find it.
# Running time: O(n). Space: O(1). Or rigorously O(lgn) given that a constant variable to represent the number "n" grows with lgn.
# http://stackoverflow.com/questions/6420467/find-any-one-of-multiple-possible-repeated-integers-in-a-list/6420503#6420503
import random
import pytest
def find_dup(lst):
assert(len(lst) >= 2)
assert(all(map(lambda x: 0 < x < len(lst), lst)))
#find cycle